2018-01-08 01:09:29 good evening 2018-01-08 01:09:55 jirutka: i hear you dont want to 2018-01-08 01:09:59 maintain rust anymore? 2018-01-08 01:10:12 leo-unglaub: not yet, but i’m very close to it 2018-01-08 01:10:30 oh nooo 2018-01-08 01:10:39 leo-unglaub: Cargo is the main reason 2018-01-08 01:10:42 hehehe 2018-01-08 01:10:51 i am starting to love cargo ... 2018-01-08 01:10:57 what is wrong with it? 2018-01-08 01:11:14 LOVE?! I’m starting to think that it’s the worst thing that happened to Rust 2018-01-08 01:11:19 basically… everything? 2018-01-08 01:11:40 but don’t worry, I’m currently working on bootstrapping Rust using https://github.com/thepowersgang/mrustc :) 2018-01-08 01:11:41 well, i am starting to love some parts of it. in simple projects its okay ... 2018-01-08 01:11:44 and it goes very well 2018-01-08 01:12:24 Cargo is bloated as hell, everything-but-kitchen-sink tool, but what’s the worst, they made it basically essential part of Rust, but it’s being developed in hipster style 2018-01-08 01:12:59 yeah, i am not a fan of all those new "ecosystems" ... but its still a lot better than npm or pip 2018-01-08 01:13:13 well, everything is better than npm XD 2018-01-08 01:13:15 but yeah, the concentration on a single point of failure is a problem 2018-01-08 01:13:44 it would be to a longer discussion; I have very bad experience with Cargo from position of package maintainer 2018-01-08 01:13:59 but I’d like to finish something now before going to sleep… 2018-01-08 01:14:04 btw will you come to FOSDEM? 2018-01-08 01:14:25 yeah sure it sucks when you compare it to the good old times ... but when you have to work with npm at work ... you see it not as that bad *g* 2018-01-08 01:14:41 nah, i never go to those conferences 2018-01-08 01:14:41 don't fall for low bars 2018-01-08 01:14:55 leo-unglaub: why not? 2018-01-08 01:15:00 "at least it's not javascript" is a very sad reason to rejoice 2018-01-08 01:15:05 leo-unglaub: what about exception this time? :) 2018-01-08 01:15:06 jirutka: i hate traveling 2018-01-08 01:15:22 leo-unglaub: come on, it’s just Brussels, one hour in plain 2018-01-08 01:15:43 i am nearly 2 meters large ... i dont fit into those modern airplanes .. 2018-01-08 01:16:04 but yeah, i should attend stuff like that more ... talk to other developers, ... 2018-01-08 01:16:04 leo-unglaub: ah, understand, planes are really not very comfortable 2018-01-08 01:16:38 leo-unglaub: but… I’ll invite you to a beer and famous brussels waffles! what about that? ;) 2018-01-08 01:16:47 i dont care about comfort ... my legs just dont fit *g* also my girlfiend beats me up again if she catches me looking at all the strewardesses 2018-01-08 01:16:56 LOL 2018-01-08 01:17:09 oh boy, i hope this was not a COC violation *g* 2018-01-08 01:17:25 I hope not! 2018-01-08 01:17:35 it was a hell good reason XD 2018-01-08 01:18:04 maybe find some plane without hot stewardesses? XD 2018-01-08 01:18:25 skarnet: what about about? FOSDEM 2018-01-08 01:21:20 what about about? 2018-01-08 01:21:33 skarnet: eh, sry, what about you and FOSDEM? 2018-01-08 01:21:40 not going this year 2018-01-08 01:21:44 hopefully next year 2018-01-08 01:21:46 skarnet: whyyy? :( 2018-01-08 01:21:52 because I have nothing new to show 2018-01-08 01:21:59 ncopa is not going, you’re not going :( 2018-01-08 01:22:12 and need to work and make $$$ to have more stuff to show next year 2018-01-08 01:22:16 that doesn’t matter, I also don’t have any talk, just to meet people from the community 2018-01-08 01:23:26 nah, I've tasted fame, now I want to be in front of the cameras or gtfo 2018-01-08 01:23:39 XD 2018-01-08 01:24:05 well, prepare some talk about whatever you want just for us? 2018-01-08 01:24:42 maybe about “slash package”, how to deal with dependencies between packages, compatibility layer and all related issues 2018-01-08 01:25:08 that would be damn interesting topic 2018-01-08 01:25:51 or about your IPC bus, if you already have some ideas about wider architecture 2018-01-08 01:27:16 or submit a talk about how awesome cargo is and piss of jirutka *g* 2018-01-08 01:27:48 leo-unglaub: eh, skarnet, how awesome hipster pkg-build-coffee-whatever-manager is, really? XD 2018-01-08 01:28:46 I'm certainly not lacking talk ideas 2018-01-08 01:29:04 I'm lacking time, money, and concrete achievements 2018-01-08 01:29:21 when I have all 3, I'll talk your heads off 2018-01-08 01:30:04 leo-unglaub: to be clear, cargo provides very good features and user comfort, the problem is technical quality etc. 2018-01-08 01:30:46 jirutka: a coworker of me has a browser adding that always closes the current tab when the words "reimagined", "redesigned", "coffee", "invisioned" are found in the current website ... 2018-01-08 01:30:59 maybe he should try to open the cargo site some time soon *g* 2018-01-08 01:31:14 skarnet: hm, i remember that you planned to quit your contract and start working on s6 revolution half year after the last FOSDEM… so what happend? 2018-01-08 01:32:04 exactly that happened, but I didn't make a full year of contracts 2018-01-08 01:32:52 also, the s6 revolution got delayed because I came to understand there's a final piece missing 2018-01-08 01:33:09 the s6 revolution will happen when I have written that final piece 2018-01-08 01:33:24 and I can't do it before making a few more contracts 2018-01-08 01:35:04 you mean proper UI? 2018-01-08 01:35:52 I wish I’d be more focused, that’s one of many many things I’d like to help with, but so little time :( 2018-01-08 01:37:50 yeah, a friendly UI with more automation and some baked-in policy. I'll handle it, I know what to do. It will just take time. 2018-01-08 01:37:58 btw, unrelated, just come to my mind… i wonder, why I can’t execute `pipeline` directly, without `execline`? it’s a normal binary, it has main(), processes arguments, but when I try to use it directly it does not work 2018-01-08 01:38:12 you can, you just need to use the right syntax 2018-01-08 01:38:39 hm, to be honest, I’d think that you’re like Linus… do you know what Linus said about him and designing UI? 2018-01-08 01:38:51 he hates it, yes. 2018-01-08 01:38:56 not just that 2018-01-08 01:39:09 he doesn't trust himself to build one. I know. 2018-01-08 01:39:19 he said that if he get into an island and the only way how to get out is to design a good UI, he would die there 2018-01-08 01:39:29 could you please give me some example how to call pipeline directly? 2018-01-08 01:39:40 A | B 2018-01-08 01:39:48 -> pipeline " A" "" B 2018-01-08 01:39:49 without { } ? 2018-01-08 01:40:24 don't write the braces. Prepend every argument inside the block with a single space. At the end of the block, add an empty argument. 2018-01-08 01:40:45 (the empty argument is used to separate both command lines) 2018-01-08 01:41:23 I’m not sure I understand, how it would look like with passing some arguments to A and B? 2018-01-08 01:41:23 foo bar corge | grault baz qux -> pipeline " foo" " bar" " corge" "" grault baz qux 2018-01-08 01:41:41 aha! 2018-01-08 01:41:50 that’s kinda… weird… why it works like that? 2018-01-08 01:42:09 <^7heo> the empty string is used as a separator? 2018-01-08 01:42:17 the empty string is a separator, yes 2018-01-08 01:42:32 <^7heo> I guess that's the lesser of all evils. 2018-01-08 01:42:41 and why one space before…? 2018-01-08 01:43:03 and the prepend thing is necessary to avoid security risks in substitutions 2018-01-08 01:43:05 <^7heo> to make the difference in the sequences? 2018-01-08 01:43:07 <^7heo> ah 2018-01-08 01:43:11 aha 2018-01-08 01:43:55 (Implementation note: the fact that word prefixes are kept is what makes execline's subtitutions secure. Blocks are implemented via prefix space characters; a substitution occurring inside a block will always produce words beginning with the right amount of spaces, thus substituted values cannot prematurely terminate a block.) 2018-01-08 01:44:10 <^7heo> ACTION is amused at how the choice of placeholder names is revealing about someone's personality 2018-01-08 01:44:42 okay, thanks! I came to that I’m not able to do that in shell… not reliably in all situations 2018-01-08 01:45:17 without prepending arguments inside a block with some characters, substitutions could throw a wrench into blocks, completely modifying the script's execution flow 2018-01-08 01:45:26 and that was bad 2018-01-08 01:45:34 <^7heo> it's funny, I wondered not so long ago how I could reliably replace the pipe in a shell by a command 2018-01-08 01:45:44 <^7heo> I didn't search for long but I found no elegant solution. 2018-01-08 01:45:49 <^7heo> I find this pretty cool. 2018-01-08 01:46:19 but seriously, if you want to avoid wasting brain space on this, just write execlineb -Pc "pipeline { A } B" 2018-01-08 01:46:43 because the brace syntax is simpler. 2018-01-08 01:46:53 <^7heo> skarnet: any reason why you didn't copy the "A | B" syntax btw? 2018-01-08 01:47:04 can't have infix Unix commands 2018-01-08 01:47:05 skarnet: this is the best I can do in POSIX shell: https://dpaste.de/7C4n … but it’s not possible to make it kill created processes when you send it KILL signal 2018-01-08 01:47:56 skarnet: and some other situations does not handle well, but I don’t remember which; the conclusion is that it’s simply not possible to emulate `pipeline` in shell 2018-01-08 01:48:04 a Unix command is: command arg1 arg2 arg3... you can't have a | in the middle of a command line and expect your first command to understand it. 2018-01-08 01:48:43 jirutka: yes, the problem with shell pipelines is that the shell has to stay around 2018-01-08 01:48:54 and this is not good when you try to supervise a pipeline 2018-01-08 01:49:33 #!/bin/execlineb -P pipeline { mdevd-netlink -v 3 } mdevd -v 3 -f /etc/mdev.conf -n 2018-01-08 01:49:34 skarnet: shell staying around is not a problem, killing subprocesses when the shell is killed is 2018-01-08 01:49:38 will do what you want 2018-01-08 01:49:47 that's exactly the problem 2018-01-08 01:50:08 you want to send signals to the daemons, not to the shell :P 2018-01-08 01:50:09 I want to modularize execline a bit and use just pipeline as standalone command :) 2018-01-08 01:51:36 this is cool, but unfortunately badly implemented https://github.com/flonatel/pipexec 2018-01-08 01:51:49 I'll never understand why you guys are so hell-bent on saving 300k of disk space when you have the whole fukken OpenRC state machine there 2018-01-08 01:52:56 if you want to extract the pipeline code from execline and include it with the mdevd package, be my guest, but it will be duplicated code at some point because s6 needs execline :P 2018-01-08 01:53:01 it’s just a waste to add 300 kiB dependency to few kiBs tool such as mdevd just for such tiny thing, moreover when you actually need just pipeline which is very small 2018-01-08 01:53:23 same could be said about most busybox utilities :P 2018-01-08 01:53:40 no, there’s one huge difference 2018-01-08 01:54:04 busybox provides many very essential utilities used by all others 2018-01-08 01:54:12 but no one use execline, sry ;) 2018-01-08 01:54:25 <^7heo> that's kind of an egg/chicken situation. 2018-01-08 01:54:29 ^ 2018-01-08 01:54:55 <^7heo> and without anyone taking the first step, it will never change. 2018-01-08 01:55:05 tbh I don’t want this change 2018-01-08 01:55:09 I don’t like execline 2018-01-08 01:55:10 I gave you a solution to your problem, if you don't want to pay the costs associated with that solution, have fun trying to hack a correct pipeline in shell 2018-01-08 01:55:17 <^7heo> yeah, that's more honest to say ;) 2018-01-08 01:55:30 nothing comes for free 2018-01-08 01:55:45 <^7heo> gmail does 2018-01-08 01:55:46 and especially not my time at 3 am 2018-01-08 01:55:48 good night 2018-01-08 01:55:48 <^7heo> ACTION hides 2018-01-08 01:55:53 <^7heo> night again, skarnet 2018-01-08 01:55:54 that said, after playing with that I kinda understand some ideas behind execline better 2018-01-08 01:56:17 <^7heo> and, crap, it's 3 AM 2018-01-08 01:56:33 <^7heo> I hate the openssl frontend. 2018-01-08 01:56:35 <^7heo> Good night. 2018-01-08 01:56:40 ok, good night 2018-01-08 01:57:27 ^7heo: and what about you and FOSDEM? 2018-01-08 01:58:50 <^7heo> it was this week end. 2018-01-08 01:59:09 no, it will be in February 2018-01-08 01:59:11 <^7heo> ah? 2018-01-08 01:59:24 3–4 February 2018-01-08 01:59:27 <^7heo> Ah 2018-01-08 01:59:30 <^7heo> I'm bad at this. 2018-01-08 02:00:59 so…? 2018-01-08 02:01:26 <^7heo> Maybe I'll come. 2018-01-08 02:01:29 <^7heo> I'm not sure yet. 2018-01-08 02:02:22 ok 2018-01-08 02:10:50 damn, have to reboot 2018-01-08 08:55:52 kaniini, thoughts on http://tpaste.us/Kdgg ? 2018-01-08 09:00:57 think you typo'd in update.c: 2018-01-08 09:00:58 + { .name = "update", .forced_force = APK_FORCE_REFRESH }, 2018-01-08 09:01:05 shouldn't that be forced_flags? 2018-01-08 09:04:40 Aerdan[m], no 2018-01-08 09:04:43 it's correct 2018-01-08 09:04:55 there's recent commit splitting the force flags 2018-01-08 09:05:02 but hmm 2018-01-08 09:05:15 dist-upgrade would be 'upgrade -a' which is applet specific flag 2018-01-08 09:06:56 this might be little bit cleaner and using less data http://tpaste.us/oQan 2018-01-08 09:09:27 it's only few applets requesting specific flags; perhaps the .parse() could be called to allow overriding the flags when alias is known 2018-01-08 09:14:31 aha, sorry. brainfart. :) 2018-01-08 09:14:42 hi all, stuipd question: if I submit a patch for an existing package to add features, how would I submit changes to the documentation? 2018-01-08 09:22:16 simonk, depends where the documentation is 2018-01-08 09:22:26 is wiki page? part of package contents? 2018-01-08 09:33:48 it currently is in the wiki only. It is about mkinitfs and keyfiles for LUKS for encrypted rootfs 2018-01-08 09:34:20 basically support for keyfiles as well as support for keyfiles on external media 2018-01-08 09:56:28 kaniini, ok, applet aliases, attempt #3 http://tpaste.us/jDXn 2018-01-08 10:00:44 as proper commit http://tpaste.us/DeLX 2018-01-08 10:00:48 off for lunch 2018-01-08 14:00:18 doing the 'defaults' options file for apk is slightly tricky due to --root... i wonder if we should enforce order of: apk [global options] [applet options] 2018-01-08 14:01:04 or maybe just rewrite the option handling by not using getopt(), and doing it ourself, so we have more control over the ordering of things 2018-01-08 14:04:30 subcommands are always a pain :/ 2018-01-08 14:05:57 having alias framework that users can use is fine, but is adding concrete aliases really a good thing? I like apk's conciseness merged with readability. apk add is a thing admin has to remember on alpine just like she or he has to remember pacman -S (I hate pacman commands) on arch, etc. 2018-01-08 14:07:38 it's perception issue mostly 2018-01-08 14:08:15 maybe i send an email about it. i want to tag a new release. perhaps without any command handling changes for now. 2018-01-08 14:08:37 dist-upgrade and purge seem like nice additions, but install and remove (why not uninstall for instance?) seem not really necessary. 2018-01-08 14:10:42 (also there is a typo in the commit message s/toi/to/) 2018-01-08 14:36:52 install and remove are offered primarily because users expect apk to respond to the subcommands they know from other package managers that aren't user-unfriendly (*coughpacmancough*) 2018-01-08 14:37:59 it takes time to override muscle-memory `$pkgman install` reflex. 2018-01-08 14:46:08 I don't deny it, but it would be like admitting that apk made a wrong choice choosing add/del, and I don't think it was a wrong one. add/del are nice, short, symmetrical. install/remove aren't. maybe install should simply spit out warning that apk has add command (instead of whole usage/help message), and remove should mention del command, but be it install or remove - they shouldn't do any 2018-01-08 14:46:14 action. it's more user-friendly, yet does not abandon original choices. and then if user cannot live w/o such aliases doing actions, then I would add possibility to define them by users 2018-01-08 14:47:47 in the end, all that doesn't matter. we want to keep the users we attract, so we have to fulfill expectations long enough for them to read the documentation about what else apk can do. 2018-01-08 15:07:08 I agree with przemoc, I expressed the same concern against aliases few days ago 2018-01-08 15:08:21 if you really want to add aliases, then I’m for install / uninstall for symmetry, install / remove really doesn’t make sense 2018-01-08 15:08:44 btw why not just add shell aliases for that? ;) 2018-01-08 15:17:15 You can't use shell aliases for apk subcommands 2018-01-08 15:17:26 There isn't a way to alias based on the second word in a command 2018-01-08 15:17:37 You'd have to make it a shell function and then parse the argument list, and that gets ugly fast 2018-01-08 15:20:04 pardis: eh, you’re right; I should rather drink me coffee first :/ 2018-01-08 15:21:59 i don't like any aliases being added to be honest... 2018-01-08 15:40:29 what about writing a standalone program just for aliases instead? let’s call it e.g. apk-noob or apk-familiar and the only thing what it would do is calling apk with proper subcommand / arguments :P unix philosophy FTW! XD 2018-01-08 15:43:21 or one step more, name it apt-get, alias all common subcommands and options and add some `sleep n` to let it really feel like apt-get; users will be happy! :P 2018-01-08 15:43:29 apktitude? 2018-01-08 15:45:41 but then another problem will appear… noobs that refuses to understand that different distros use different names for some packages… what next? is someone gonna propose to add gazillion provides= as alternative names with these poor debian naming conventions (gazillion of pkgs libxxx)? 2018-01-08 15:49:43 Let apk be apk. Every distro has its own peculiarities, like sillyness having to use a differemt command for searching versus installing, or naming things libXXX-dev versus libXXX-devel. 2018-01-08 15:50:07 to end this rant, I’m really afraid that this is not a good way and there’s not a clear boundary where to stop with trying to be more familiar for users of other distros; if it’s not gonna add any complexity or just very little, then fine, but is this that case? is this abstraction you’re building for aliases actually useful even for some other use cases than this? 2018-01-08 15:53:37 as I’m thinking about it… git has very good support for aliases, it’s configurable from .gitconfig and quite flexible… why not to implement similar mechanism as a standalone program that can be used to make aliases for any other program, not just apk? 2018-01-08 15:54:39 so just make symlink e.g. aliases -> apk and add `apk install = apk add` to some configuration file 2018-01-08 15:57:14 hm, not so good example with git, it seems that it doesn’t allow to alias arguments, just subcommands; but it should not be a problem to implement such functionality 2018-01-08 18:24:29 przemoc: i do think you have a point in that a simple alias to install/remove may be misleading, since apk add/del does not actually add or remove a package (this happens by side effect); but i think adding an alias is still valid 2018-01-08 18:37:09 usability is valuable 2018-01-08 18:41:29 we are not adding help for those anyways... though that might cause questions later on 2018-01-08 19:25:47 kaniini: this is more about familiarity than usability 2018-01-08 19:26:09 kaniini: simple vs. easy (familiar) 2018-01-08 19:26:33 kaniini: Alpine puts attention on simplicity, not familiarity 2018-01-08 19:27:28 kaniini: this is great talk from Rich Hickey on this topic https://www.youtube.com/watch?v=rI8tNMsozo0 2018-01-08 19:30:50 Ruby is quite known for its “more ways to do one thing”, specifically many “aliases” in stdlib, i.e. one function may have multiple names, e.g. reduce and inject, map and collect etc. 2018-01-08 19:32:35 it may look fine, anyone can use his/her preferred naming based on his/her background, but I can tell you, it’s not helping at all to newbies; it’s nothing but complication when you teach Ruby and have to explain students that “these two things are actually the same, just different names” all over, it’s very confusing for them 2018-01-08 19:33:07 so you see a function you don’t know… is it a new function? or is it just different name for a function I already know? 2018-01-08 19:33:26 does inject and reduce behaves the some or is there some difference? 2018-01-08 19:34:05 what the hack is “add” in apk? how is that different from “install”? do they do the same or not? 2018-01-08 19:35:22 aliases are fine when *user* defines them for themselves, not when predefined by the author of software; this just leads to confusions, unclear manuals etc. 2018-01-08 19:37:08 kaniini: also, did you read my suggestion to write generic utility for aliases? WDYT? 2018-01-08 20:04:35 jirutka: i think that i want to do something to remove a pain point that has tons of actual data to support it being a pain point 2018-01-08 20:04:44 jirutka: i don't really care the mechanism used 2018-01-08 20:05:33 you don't understand the problem, because you're approaching it from the position of bias as a seasoned alpine user/dev 2018-01-08 20:05:52 I understand the problem very well 2018-01-08 20:06:08 yes, I’m obviously biased, but I understand background of this 2018-01-08 20:06:35 alpine puts attention on simplicity because simplicity is usually an enlightening path to usability 2018-01-08 20:06:51 heh, no XD 2018-01-08 20:06:58 it depends 2018-01-08 20:07:03 it depends on target group 2018-01-08 20:07:56 it really doesn't. i was actually in the IRC channel when we literally came up with the mantra of "small, simple, secure" 2018-01-08 20:08:02 it is a usability statement 2018-01-08 20:08:14 no, it surely does 2018-01-08 20:08:35 Alpine is simply for power users 2018-01-08 20:08:45 Go Go Power Users! 2018-01-08 20:08:46 it’s not simply for uneducated newbies 2018-01-08 20:08:50 simple 2018-01-08 20:09:54 Alpine is for anybody who wants to use Alpine 2018-01-08 20:10:44 you’re really confusing terms 2018-01-08 20:12:16 usability is strongly related to audience 2018-01-08 20:12:35 CLI is very usable for sysadmins, but not for secretary 2018-01-08 20:12:45 fancy GUI is very usable for secretary, but hell not for sysadmins 2018-01-08 20:13:00 CLI is basically always more simple in terms of complexity than GUI 2018-01-08 20:13:26 it’s total bullshit to say that Alpine should be easy for every user 2018-01-08 20:13:40 Hi. As a disabled user, I would like to point out that engineering towards the disadvantaged (i.e. those who have less experience, in this case) generally makes software and environments more friendly to those who aren't disadvantaged (i.e. sysadmins and engineers and so on and so forth). 2018-01-08 20:14:07 No, it's total bullshit to say that Alpine should intentionally be more difficult to use than it needs to be. 2018-01-08 20:14:25 usability does also involve CLIs 2018-01-08 20:14:52 and, while `apk add/del` are right, it is at least useful to have a pointer with `apk install/remove` 2018-01-08 20:15:02 it increases complexity 2018-01-08 20:15:04 by the way, the whole concept of apk working on a ruleset is also an intentional usability decision 2018-01-08 20:15:06 so it’s not simpler 2018-01-08 20:15:18 it’s easy, familiar for former debian/whatever users 2018-01-08 20:15:28 How does it increase complexity to have install/remove aliased to add/del? 2018-01-08 20:15:54 "No, it's total bullshit to say that Alpine should intentionally be more difficult to use than it needs to be." … I didn’t say anything like that, this is wrong conclusion 2018-01-08 20:16:03 jirutka: i don't care if it is an alias or not 2018-01-08 20:16:03 more lines of code to maintain 2018-01-08 20:16:13 if `apk install` prints "use apk add you idiot" 2018-01-08 20:16:26 then that's fine too 2018-01-08 20:16:37 i mean, i would prefer better wording than that, but you get the point 2018-01-08 20:17:05 and not just more lines of actual code, even in documentation 2018-01-08 20:17:08 the problem is that when you do `apk install` as you would expect based on the Debian and RPM family distributions to work 2018-01-08 20:17:21 and you get presented with pages of help text 2018-01-08 20:17:33 it wastes mental bandwidth 2018-01-08 20:17:47 and people do give up on trying alpine as soon as they encounter that 2018-01-08 20:17:52 and then they say "apk sucks" 2018-01-08 20:18:15 when you do `apt-get install foo`, you would expect based on the Debian distributions to work 2018-01-08 20:18:36 they never give it a fair chance because it doesn't even provide useful advice for them trying to install a package 2018-01-08 20:18:38 apk has a lot of usage information spewd. It would be far more helpful to declare that the command offered was incorrect and direct people to a well-written manpage which documents the application. Or at least offers useful pointers like "did you mean 'apk add'?" 2018-01-08 20:18:48 for that they have to read through the help message and it's many lines down 2018-01-08 20:18:49 sorry, if user give up on Alpine b/c `apk install` prints help instead of install package, then it’s users fault, not ours 2018-01-08 20:18:54 this is utter stupidity 2018-01-08 20:19:39 it's really not 2018-01-08 20:19:49 Yes, I agree. Alpine should just ship a telepathy module so that it knows precisely what the user wants. /s 2018-01-08 20:19:56 ^ +1 2018-01-08 20:19:59 and that type of mindset will only hold alpine and the overall design philosophy behind alpine back 2018-01-08 20:20:14 no, YOUR mindset leads directly to Ubuntu hell 2018-01-08 20:20:35 In the real world, apk and other Alpine tools need to be user-friendly. This means behaving in a way which encourages users to keep using the software. 2018-01-08 20:21:02 This means that 'apk install' needs to exist. Whether it does the right thing from the user's perspective or directs their attention to 'apk add' is immaterial. 2018-01-08 20:21:04 but I agree that `apk install` should probably print "Unknown command" or something like that followed by help 2018-01-08 20:21:06 jirutka: my mindset is "do something more reasonable than spew 40 lines of usage information" when the user does `apk install` 2018-01-08 20:21:10 curretly it prints only help 2018-01-08 20:21:28 Printing help does not help. 2018-01-08 20:21:32 why does not? 2018-01-08 20:21:50 Forty lines of crap is a lot to wade through when you just want it to DO THE THING. 2018-01-08 20:21:50 7th line: add Add PACKAGEs to 'world' and install (or upgrade) them, while ensuring that all dependencies are met 2018-01-08 20:22:04 correct 2018-01-08 20:22:06 okay, let’s restruture that help 2018-01-08 20:22:11 so what i am saying is 2018-01-08 20:22:11 instead of print help 2018-01-08 20:22:13 it’s not crap 2018-01-08 20:22:19 what we could do 2018-01-08 20:22:22 if we don't want an alias 2018-01-08 20:22:24 is have it like this 2018-01-08 20:22:30 $ apk install 2018-01-08 20:22:38 Oh. Correction. It's *ninety-nine* lines of crap. 2018-01-08 20:22:42 kaniini: did you ever study UI/UX? 2018-01-08 20:22:43 apk: unknown command 'install', did you mean 'apk add'? 2018-01-08 20:22:50 psychology of users? 2018-01-08 20:22:53 and related topics? 2018-01-08 20:22:59 So much crap that the user has to rerun the command through a pager to even *see* 'add'. 2018-01-08 20:23:03 jirutka: yes, i actually have a degree in it 2018-01-08 20:23:29 Guys 2018-01-08 20:23:30 Guys 2018-01-08 20:23:35 man apk 2018-01-08 20:23:36 Please 2018-01-08 20:23:37 Do it 2018-01-08 20:23:49 consus: it is on my todo list 2018-01-08 20:24:02 Telling people to RTFM is not conducive to getting them to stick with us. ;) 2018-01-08 20:24:10 Err 2018-01-08 20:24:24 lsblk: invalid option -- 'Z' 2018-01-08 20:24:24 Try 'lsblk --help' for more information. 2018-01-08 20:24:38 Replace lsblk --help with man apk and it's okay 2018-01-08 20:24:45 kaniini: Aerdan[m]: less crap in help: https://dpaste.de/r7LH/raw 2018-01-08 20:25:19 apk version and arch is the least important, this is usually at bottom 2018-01-08 20:25:20 jirutka: please, do not print help on errors 2018-01-08 20:25:21 I think you're intentionally missing the point. 2018-01-08 20:25:31 jirutka: print the error 2018-01-08 20:25:38 jirutka: unfortunately we have to keep the global options docs 2018-01-08 20:25:41 Most users are going to 'apk install', see that they didn't get what they want, and go "wow this distro sucks, why are we using it again?" 2018-01-08 20:25:45 this is acutally about 'printing help on errors' yes 2018-01-08 20:25:48 horribly long line with all options is totally useless 2018-01-08 20:26:22 if 'apk install' is not an alias, having a more useful message than help also solves the problem 2018-01-08 20:26:31 for example 2018-01-08 20:26:35 +1 2018-01-08 20:26:40 ACTION sent a long message: kaniini_2018-01-08_20:26:39.txt  2018-01-08 20:26:43 $ tmux unknown-command 2018-01-08 20:26:50 consus: yes, as I said, I agree that it should print error, e.g. https://dpaste.de/Vx7s/raw (but in better english) 2018-01-08 20:26:51 unknown command: unkonwn-command 2018-01-08 20:27:01 Without usage 2018-01-08 20:27:16 printing help should only occur on --help, when it's explicitly requested. otherwies print a simple error message and suggest what people might prefer. 2018-01-08 20:27:23 Because on some consoles I cannot even see the error itself 2018-01-08 20:27:28 Only the tail of the help 2018-01-08 20:27:51 again, whether we do what the user intended or guide them to the right command is immaterial. what matters is that we need to do *something* to make apk more user-friendly. 2018-01-08 20:28:01 the two options given are the most helpful for this purpose. 2018-01-08 20:28:15 Aerdan[m]: in terms of usability for ordinal users printing help by default is surely better than printing just a error message… noob may not know that (s)he can use -h or --help… and also the problem that not all programs respect this convention 2018-01-08 20:28:19 I really don't give a shit which is chosen. I really don't give a shit whether doing the right thing is more complex. 2018-01-08 20:28:26 "Most users are going to 'apk install', see that they didn't get what they want, and go "wow this distro sucks, why are we using it again?" … and I’m telling you that when you accept this game, you lost 2018-01-08 20:28:31 apk is fine 2018-01-08 20:28:35 <_ikke_> jirutka: nod 2018-01-08 20:28:48 if user give up just because he use program he don’t know as very different program he know, then it’s his problem 2018-01-08 20:29:02 Well generally speaking: yes 2018-01-08 20:29:04 should vim add shortcut for quit from nano etc? 2018-01-08 20:29:17 jirutka: but that's not my point, my point is at least pointing them in the right direction 2018-01-08 20:29:18 this is so bullshit that I’m shocked you’re even talking about this 2018-01-08 20:29:32 obviously not 2018-01-08 20:29:32 But it would nice to a) report the error b) do not report the whole help to the degree when a user can no longer see the error 2018-01-08 20:29:39 kaniini: yes, error message and help when on the 3rd line is the command they need 2018-01-08 20:29:56 i'm talking about "handling 'apk install' in a more friendly way" 2018-01-08 20:30:13 people wont read it 2018-01-08 20:30:19 you get 2 lines max, really 2018-01-08 20:30:26 if not printing help by default, then it should at least include something like "run `apk --help` to get help" in the error message 2018-01-08 20:30:30 Yes 2018-01-08 20:30:39 I posted the lsblk output 2018-01-08 20:30:46 correct 2018-01-08 20:30:48 so 2018-01-08 20:30:52 that leaves us with 2018-01-08 20:30:53 I convinced the author to stop posting the whole help ^^ 2018-01-08 20:30:55 otherwise i can guarantee you that noobs start shouting that apk sucks because it does not even print help and they don’t know how to do that and OBVIOUSLY it’s our fault, not their 2018-01-08 20:30:55 jirutka, the whole point is that the user does not see the error message. the user does not see the proper command. they just see a bunch of crap (99 lines overflows even the 80-line terminal, may I point out). this discourages them from proceeding further. even professionals fall foul of this. 2018-01-08 20:31:51 ACTION sent a long message: kaniini_2018-01-08_20:31:50.txt  2018-01-08 20:32:06 Yeah, something like that is okay 2018-01-08 20:32:10 kaniini: this is good 2018-01-08 20:32:19 Sweet ^^ 2018-01-08 20:32:30 okay then, that's an ok solution to me 2018-01-08 20:32:38 we can search the help descriptions for the applet name given 2018-01-08 20:32:41 to see if we can find the right command 2018-01-08 20:32:58 i would still like to see configurable aliases 2018-01-08 20:33:15 I would still like to see a water fountain in apk 2018-01-08 20:33:18 :D 2018-01-08 20:33:20 Btw 2018-01-08 20:33:22 and it definitely should control my coffee machine!! 2018-01-08 20:33:31 About apk and configuration options 2018-01-08 20:33:37 jirutka: apk fetch -vvvvv coffee 2018-01-08 20:33:42 I think install docs package to get the man pages is kinda odd 2018-01-08 20:33:46 "add Add PACKAGEs to 'world' and install (or upgrade) them, while ensuring that all dependencies are met" … this is technically correct, but maybe it’d be better to somehow mention "install" at the start 2018-01-08 20:33:50 oops 2018-01-08 20:33:55 jirutka: apk fetch --force coffee 2018-01-08 20:34:05 It would be better to have a /etc/apk/apk.conf of some sort 2018-01-08 20:34:38 consus: we are adding /etc/apk/defaults 2018-01-08 20:34:39 Because there is literally NO mentioning of this magic aside from the apk sources itself 2018-01-08 20:34:55 consus: as I said before, aliases for subcommand and arguments is something that can be generalized to a standalone lightweight program with configuration file that can be used for anything, not just apk 2018-01-08 20:35:08 jirutka: agreed 2018-01-08 20:35:11 consus: modularity, unix philosophy 2018-01-08 20:35:30 consus: what magic do you mean? 2018-01-08 20:35:32 jirutka: Though configuration options like 'install/do not install' man pages should go to the config 2018-01-08 20:35:35 apk add docs 2018-01-08 20:35:40 SHAZAM 2018-01-08 20:35:45 Now I have man pages 2018-01-08 20:35:53 aha, this 2018-01-08 20:36:00 Not because docs is a virtual 2018-01-08 20:36:03 well, this is not about apk, more about aports 2018-01-08 20:36:09 But because apk tests for it 2018-01-08 20:36:09 and yes, it should be better documented 2018-01-08 20:36:15 Yeah 2018-01-08 20:36:21 That's why I registered on wiki 2018-01-08 20:36:23 Got banned 2018-01-08 20:36:33 And started a thread to redo documentation xD 2018-01-08 20:36:48 BTW I'm back from vacanation so it's time to write some posts 2018-01-08 20:37:10 banned? did someone already unbanned you? 2018-01-08 20:37:14 Yeah 2018-01-08 20:37:17 I suppose 2018-01-08 20:37:26 I asked where should this kinda docs go 2018-01-08 20:37:30 In which section I mean 2018-01-08 20:37:41 banned from what 2018-01-08 20:37:48 And now there is a thread on ML called 'What the hell should we do with the Alpine documentation' 2018-01-08 20:37:51 From Alpine Wiki 2018-01-08 20:38:11 i think the wiki should be burned to the ground personally 2018-01-08 20:38:16 Well 2018-01-08 20:38:28 i would prefer to see something like freebsd handbook for alpine 2018-01-08 20:38:28 I think not 2018-01-08 20:38:31 So there is a thread 2018-01-08 20:38:40 'Improving Alpine Linux documentation' 2018-01-08 20:38:43 Dec 12 2018-01-08 20:38:55 If you have something to say -- post it there 2018-01-08 20:39:13 It's time to come up with a decision I gues 2018-01-08 20:40:35 the main problem with the current wiki is that people often don’t update it… and this includes even kaniini, you added new feature to apk but did not add documentation for it to wiki… so i copied it from your arising documentation when noticed it 2018-01-08 20:40:53 There is my point exactly 2018-01-08 20:41:03 Well, that apk should be documented in man pages 2018-01-08 20:41:05 Not on wiki 2018-01-08 20:41:08 yes 2018-01-08 20:41:11 Because wiki is for other things 2018-01-08 20:41:20 i am working on that 2018-01-08 20:41:28 Like guidelines, or 'how the hell should I start my nginx' 2018-01-08 20:41:33 the problem is not “wiki”, the problem is Mediawiki 2018-01-08 20:41:42 Please, post it there guys 2018-01-08 20:41:55 Let's come up with something already 2018-01-08 20:41:59 we plan to produce a manual for Adélie too 2018-01-08 20:42:07 And I'll finally start some doing instead of talking 2018-01-08 20:42:11 which would largely be reusable for upstream Alpine 2018-01-08 20:42:22 but the $$$ isn't there for a doc writer yet 2018-01-08 20:42:24 need to concentrate on code 2018-01-08 20:43:39 and since people donjt update current wiki, we don’t have just technological problem (MediaWiki), but also problem with missing/non-quality content 2018-01-08 20:44:21 Yeah 2018-01-08 20:44:55 jirutka: I have a spare time today, do you want a man page for apk? 2018-01-08 20:44:56 a problem with documenting apk right now is 2018-01-08 20:44:59 jirutka: I can write it 2018-01-08 20:45:09 In mdoc 2018-01-08 20:45:11 we want to redesign some of the applets 2018-01-08 20:45:12 AFAIR 2018-01-08 20:45:20 which were done quickly 2018-01-08 20:45:23 consus: i will handle the apk manpage 2018-01-08 20:45:26 consus: uh, could you write it in something *people* can work with? 2018-01-08 20:45:31 xD 2018-01-08 20:45:35 mdoc looks more like a machine format 2018-01-08 20:45:38 AFAIR you guys wanted mdoc 2018-01-08 20:45:45 kaniini wanted mdoc 2018-01-08 20:45:45 consus: but would appreciate reviews 2018-01-08 20:45:47 tbh, wikis are a terrible choice if there's no dedicated staff to keep it fresh and accurate. 2018-01-08 20:45:56 yes, i want mdoc 2018-01-08 20:46:10 Well, I wrote in mdoc for my soft 2018-01-08 20:46:16 Pretty okay 2018-01-08 20:46:22 Better than groff 2018-01-08 20:46:27 kaniini: then don’t expect many people to contribute into it… you will be on your own, maybe with consus 2018-01-08 20:46:38 b/c not many people are willing to work with such insanly horrible format 2018-01-08 20:47:00 The other option is something like asciidoctor 2018-01-08 20:47:04 jirutka: i think technical docs concerning apk should be maintained by the people who actually know how apk works internally 2018-01-08 20:47:08 Git / CGit uses it 2018-01-08 20:47:28 the manpage isn't really intended to be edited by lots of people, and requiring custom software just to translate the markup language du jour to mdoc so that it can be rendered properly by tha man backend is just bad design. 2018-01-08 20:47:31 i do not want to add asciidoctor as build dependency for apk 2018-01-08 20:47:42 yeah, I vote for Asciidoctor; it has builtin support for man pages; e.g. this is AsciiDoc https://github.com/jirutka/esh/blob/master/esh.1.adoc 2018-01-08 20:47:48 and I generate a standard man page from it 2018-01-08 20:47:53 sigh 2018-01-08 20:47:58 But yeah 2018-01-08 20:48:09 Having asciidoctor as a build dep is something not very okay 2018-01-08 20:48:13 Cause it wants ruby 2018-01-08 20:48:16 yeah 2018-01-08 20:48:17 And all other kind of stuff 2018-01-08 20:48:18 since this is going to be a shitshow 2018-01-08 20:48:21 I vote for "let kaniini, the person who *designed apk to begin with*, do their job" 2018-01-08 20:48:24 i am going to just drop everything 2018-01-08 20:48:28 and write this manpage 2018-01-08 20:48:35 but it’s just to built a docs, so it’s not a big deal IMO 2018-01-08 20:48:36 xD 2018-01-08 20:48:46 Aerdan: well not all of it, but many elements of the UX 2018-01-08 20:49:04 for example how pinning works :) 2018-01-08 20:49:07 but if kaniini will write the docs for apk-tools and will maintain it in long-term, then it’s fine for me 2018-01-08 20:49:22 kaniini: If you don't have time I can help you, not a big deal 2018-01-08 20:49:57 consus: i already have a decent mental map of how i want it to be 2018-01-08 20:50:07 single or per-applet? 2018-01-08 20:50:19 one of many benefits of writing it in AsciiDoc is that you can use the same content for more than just a unix man page 2018-01-08 20:50:28 jirutka: behold 2018-01-08 20:50:37 I mean really 2018-01-08 20:50:40 It is very funny 2018-01-08 20:50:41 But 2018-01-08 20:50:46 OpenBSD folks write a converter 2018-01-08 20:50:54 FROM mdoc to markdown 2018-01-08 20:50:54 xD 2018-01-08 20:51:00 jirutka: mdoc can render to html 2018-01-08 20:51:05 And html, right 2018-01-08 20:51:10 but that said, I kinda understand that it’s not very good to have asciidoctor and Ruby as dependency for apk even just for building docs 2018-01-08 20:51:18 http://man.openbsd.org/dmesg 2018-01-08 20:51:21 For example 2018-01-08 20:52:04 The css is ugly as hell 2018-01-08 20:52:08 But it's just css 2018-01-08 20:52:37 the markup (HTML) is shit too 2018-01-08 20:52:44 Why? 2018-01-08 20:52:51 just look at it 2018-01-08 20:52:54 It's clean and easy to read 2018-01-08 20:52:58 Ah 2018-01-08 20:53:00 HTML 2018-01-08 20:53:02 it is pretty trivial to write a script that converts roff/mdoc to html. mdoc even provides one. 2018-01-08 20:53:06 Who cares 2018-01-08 20:53:13 but well, if you don’t use my real HTML5 convertor for Asciidoctor, then built-in Asciidoctor output is even worse XD 2018-01-08 20:53:15 or to any other preferred presentation format, really. 2018-01-08 20:53:16 It's the browser job 2018-01-08 20:53:17 asciidoc(tor) is really not necessary. 2018-01-08 20:53:19 next you people will clamour for docs to be written in TeX. 2018-01-08 20:53:29 Notion 2018-01-08 20:53:34 Notion has docs in latex 2018-01-08 20:53:35 xD 2018-01-08 20:55:26 anyway, roff/mdoc is well-documented and straightforward. there's really no need to use another markup language just because you find that other language easier to write. 2018-01-08 20:55:34 Well 2018-01-08 20:55:45 I would REALLY like to have a simple something-other -> mdoc converter 2018-01-08 20:55:47 Belive me 2018-01-08 20:55:48 it is very "well let's just write it in JS so everyone can work on it"-tier 2018-01-08 20:55:59 Because roff/mdoc is UGLY AS FUCKING HELL 2018-01-08 20:56:11 consus: if you're serious about writing manpage and want to take a lead on it 2018-01-08 20:56:12 yeah, definitely, because writing content is the least important thing for documentation, right? oh well… XD 2018-01-08 20:56:32 consus: apt(8) on recent debian is kind of the format i had in mind 2018-01-08 20:56:41 consus: also keep in mind the apk design mantra: "if in doubt, do the exact opposite of whatever pacman does" 2018-01-08 20:56:50 XD 2018-01-08 20:56:53 :D 2018-01-08 20:57:27 kaniini: apt(8) just points you to the per-applet page, right? 2018-01-08 20:58:03 yes in part 2018-01-08 20:59:07 basically apt(8) provides a top-level overview of the new apt frontend 2018-01-08 20:59:30 and points to manpages for the old commands which provide similar functions 2018-01-08 21:01:31 Well, the thing is apk is not that big so it maybe better to put everything in one place, eh? 2018-01-08 21:02:05 I'm with either option 2018-01-08 21:03:26 i think splitting it gives more opportunity for technical detail 2018-01-08 21:03:45 but would like to discuss the world ruleset in some way in the top-level docs too 2018-01-08 21:03:54 as well as pinning 2018-01-08 21:04:30 As well as repo file format 2018-01-08 21:04:48 Okay, I got it 2018-01-08 21:05:17 the repo format probably belongs in apkindex(8), actually, but yeah. 2018-01-08 21:05:30 Nah, I mean /etc/apk/repositories 2018-01-08 21:06:02 And other files that apk processes 2018-01-08 21:06:17 Like /etc/apk/world 2018-01-08 21:06:29 Because for now we this kind of information lives on the wiki 2018-01-08 21:22:07 yes 2018-01-08 21:27:55 jirutka: anyway, i would consider ubuntu to be an example of 'bad design' in most cases. but that does not invalidate the case studies and other research they have done. it is possible to look at data and make a different conclusion. 2018-01-08 21:29:23 ubuntu's tools are condescending in terms of how they view the intellect of the user 2018-01-08 21:29:37 i do not want alpine's tools to assume the user is an idiot 2018-01-08 21:29:58 that is not, and should not be an alpine design philosophy 2018-01-08 21:33:48 ^ thank you for that 2018-01-08 21:41:37 i just want new users to feel comfortable with the tools 2018-01-08 21:42:01 this does not mean reinventing the world from the perspective that the user is an idiot 2018-01-08 21:49:26 BTW 2018-01-08 21:49:27 Guys 2018-01-08 21:49:36 Do we really need to support apk add -V syntax? 2018-01-08 21:50:32 we don't 2018-01-08 21:50:38 I mean it's hard to guess what will happen if I call apk info -i 2018-01-08 21:51:05 it just happens to be that way because of the way getopt works 2018-01-08 21:51:07 What kind of interactive output can info possibly produce 2018-01-08 21:51:12 Huh? 2018-01-08 21:51:20 fabled was talking about reworking it 2018-01-08 21:51:21 100% agree on it 2018-01-08 21:51:36 apk info --interactive wont do anything 2018-01-08 21:51:38 it's not worth documenting :P 2018-01-08 21:51:48 That's my point 2018-01-08 21:51:58 apk info -h show that there is -i switch 2018-01-08 21:52:03 And I go wow 2018-01-08 21:52:04 What 2018-01-08 21:52:05 yeah, don't document it there 2018-01-08 21:52:08 :D 2018-01-08 21:52:16 About getopt 2018-01-08 21:52:20 yes, it's a global switch 2018-01-08 21:52:27 not all global switches apply to every applet 2018-01-08 21:52:34 just enough fo them 2018-01-08 21:52:36 It's entirely possible to do apk --your-flag do-cmd -z 2018-01-08 21:52:50 yes 2018-01-08 21:52:51 i know 2018-01-08 21:52:54 Ok 2018-01-08 21:53:01 what i am saying is 2018-01-08 21:53:15 --interactive is a global switch because enough applets make use of it 2018-01-08 21:53:48 Perhaps it would be better to do it git way 2018-01-08 21:54:12 But that's another story 2018-01-09 01:39:53 jirutka: http://tpaste.us/O41Z help cleanup 2018-01-09 02:22:07 http://tpaste.us/w7j9 2018-01-09 05:56:52 http://git.infradead.org/users/dwmw2/gcc-retpoline.git/shortlog/refs/heads/gcc-7_2_0-retpoline-20171219 2018-01-09 05:57:00 this may be worth looking at for our gcc 2018-01-09 06:08:24 does the alpine kernel in current release support emmc? i forget if i've asked already 2018-01-09 06:12:38 oh, it's a module 2018-01-09 08:46:16 Xe: yes it does but you need to add it to initramfs to boot from it. 2018-01-09 08:46:31 is that done automatically? 2018-01-09 08:46:49 no 2018-01-09 08:47:38 https://git.alpinelinux.org/cgit/mkinitfs/tree/mkinitfs.conf 2018-01-09 10:06:36 is someone here able and has some time to help me debug why the testing/motion package segfaults on alpine but not on other operating systems like debian, centos, etc.? 2018-01-09 10:09:28 I don't have time and I'm probably not able, but without even looking, I can tell there's a good chance it has something to do with thread stack size. :P 2018-01-09 10:10:18 (i.e. it allocates a significant amount of stack in a thread without calling pthread_attr_setstacksize() first.) 2018-01-09 10:11:13 it says this in dmesg at the same time: "[100692.415299] grsec: From 192.168.100.1: denied resource overstep by requesting 4096 for RLIMIT_CORE against limit 0 for /usr/bin/motion[motion:17740] uid/euid:0/0 gid/egid:0/0, parent /bin/busybox[ash:17735] uid/euid:0/0 gid/egid:0/0" 2018-01-09 10:11:45 ah, I was wrong, this is grsec and not thread stack size :D 2018-01-09 10:11:57 (which was the other likely possibility) 2018-01-09 10:12:10 is it working on alpine-vanilla? 2018-01-09 10:12:23 gdb, valgrind and strace would probably be a good start for this but they all show irrelevant information and gdb only shows "??" as the location 2018-01-09 10:12:41 skarnet: i'm setting up a vanilla instance right now to see if that causes it 2018-01-09 10:13:46 the dmesg is saying motion wants to be able to dump 4k cores, which... doesn't make much sense 2018-01-09 10:14:03 and grsec prevents that, it enforces the non-dumping of cores 2018-01-09 10:14:28 requesting 4kB cores is extremely weird 2018-01-09 10:14:47 well it does that after crashing 2018-01-09 10:14:54 [100692.415113] motion[17740]: segfault at 717662daba5c ip 0000006445ddb8a3 sp 0000717662daba40 error 6 in motion[6445dc2000+3c000] 2018-01-09 10:14:56 [100692.415129] grsec: From 192.168.100.1: Segmentation fault occurred at 0000717662daba5c in /usr/bin/motion[motion:17740] uid/euid:0/0 gid/egid:0/0, parent /bin/busybox[ash:17735] uid/euid:0/0 gid/egid:0/0 2018-01-09 10:14:57 ah 2018-01-09 10:14:58 thet's what is printed before 2018-01-09 10:15:04 sorry i only copied one line 2018-01-09 10:15:06 okay, that makes more sense 2018-01-09 10:15:37 that segfault is the real issue 2018-01-09 10:16:15 keep trying with linux-vanilla just in case, but I'm coming back to my thread stack size suspicion. 2018-01-09 10:34:00 same on vanilla :x 2018-01-09 10:39:27 and gdb still says "(gdb) bt -> #0 0x000055555556d8a3 in ?? () #1 0x000055555556e98e in ?? () #2 0x000055555555fd07 in ?? () #3 0x00007ffff7dc3ff8 in ?? () from /lib/ld-musl-x86_64.so.1 #4 0x0000000000000000 in ?? ()" so i guess i'll need to build a debug version of this software 2018-01-09 10:44:07 valgrind shows this: 2018-01-09 10:44:08 ==2327== Process terminating with default action of signal 11 (SIGSEGV) Bad permissions for mapped region at address 0x408AA5C at 0x1218A3: ??? (in /usr/bin/motion) 2018-01-09 10:44:10 ==2327== Process terminating with default action of signal 11 (SIGSEGV) Bad permissions for mapped region at address 0x408AA10 at 0x4A8D609: _vgnU_freeres (vg_preloaded.c:83) 2018-01-09 10:50:23 yeah. Core devs should take over from here, but chances are it's thread stack size. If it is, it should be easy to fix. 2018-01-09 10:52:47 what do you suggest? should i contact the maintainers listed in the APKBUILD or contact the origin project on github? 2018-01-09 11:05:04 contact the maintainers first, they're probably reading this channel. They have experience dealing with porting stuff to musl and the problems that may arise. 2018-01-09 11:05:18 Then the fix can be upstreamed. 2018-01-09 11:11:49 huh, looks like building 4.1.1 instead of the currently available apk 4.0.1 fixes this issue 2018-01-09 11:15:01 Fusl, default stack size assumption is one of the most common things we need to fix when porting to musl 2018-01-09 13:51:27 this morning after rebooting my laptop wifi failed due to linux-firmware patch 2018-01-09 13:51:53 was unable to find iwlwifi-2030 ucode 2018-01-09 13:52:07 actually in /lib/firmware there were not the fw 2018-01-09 13:52:36 linux-firmware-other (which is the package containing that fw) was installed 2018-01-09 13:52:50 I had to apk fix linux-firmware-other to get the files back 2018-01-09 13:52:53 ...so... 2018-01-09 13:52:55 heads up. 2018-01-09 14:41:19 hi there, first time building packages for alpine. I have a problem with untrusted package 2018-01-09 14:42:22 on machine A, I ran abuild-keygen -a -i, build package, put resulting files to web 2018-01-09 14:42:38 server. then I did apk update and all was well. 2018-01-09 14:43:19 switched to fresh machine B (docker container), placed pub key from A in /etc/apk/keys, 2018-01-09 14:43:37 edited /etc/apk/repositories with same URL, did apk update 2018-01-09 14:43:53 now I get: 2018-01-09 14:43:54 fetch http://build.rsyslog.com/alpine/x86_64/APKINDEX.tar.gz ERROR: http://build.rsyslog.com/alpine: UNTRUSTED signature 2018-01-09 14:44:23 Can someone lend me a hand of what I am overlooking? 2018-01-09 15:01:50 oh damn, I think I found it 2018-01-09 15:06:59 copy&paste was not my friend... ;-) 2018-01-09 15:11:30 Since this is -devel, I'll rephrase my question from the general channel to post it here. Trying to build HHVM, libressl is reporting OPENSSL_MAJOR_VERSION as '2', which fails a part of the HHVM build. I can't install both openssl and libressl, so how might I get around this? 2018-01-09 15:12:59 hi. what we normally do in those cases is that we 1) report the build failure upstream, 2) create a patch for it 2018-01-09 15:14:18 you can look how we have worked around it in other packages: grep LIBRESSL_VERSION_NUMBER /.patch 2018-01-09 15:14:30 you can look how we have worked around it in other packages: grep LIBRESSL_VERSION_NUMBER */*.patch 2018-01-09 15:24:07 ncopa: thanks, I think that solved it. For logging purposes... Alpine example: https://git.alpinelinux.org/cgit/aports/tree/community/qt5-qtbase/libressl-compat.patch along with the upstream HHVM code causing problems https://github.com/facebook/mysql-5.6/blob/d26efe0d7763b6c9ba67a5cb42faebf075298b02/cmake/ssl.cmake#L106 2018-01-09 15:32:05 jyw: thanks! could you report it upstream? 2018-01-09 15:39:57 kaniini, linux-firmware-other is broke 2018-01-09 15:40:01 goes to wrong dir 2018-01-09 15:40:51 fixing that 2018-01-09 15:40:57 i think the linux-firmware split merge was too early 2018-01-09 15:41:07 it causes confusion as loads of new stuff 2018-01-09 15:41:19 yet, linux-firmware is pulled in by kernels and it pulls in all packages always 2018-01-09 15:41:23 so nothing is really achieved yet 2018-01-09 15:42:07 i'm sort of tempted to revert it 2018-01-09 15:44:30 ncopa, kaniini : ^ 2018-01-09 15:46:43 I it's easier to move forward from already splitted stuff, but if it doesn't preserve status quo atm, i.e. is partially broken, revert seems perfectly fine. it can be always applied already fixed in future. 2018-01-09 15:46:52 s/^I // 2018-01-09 15:47:08 s/already/again/ 2018-01-09 15:47:16 i'll just fix the -other package now 2018-01-09 15:47:25 that should btw. split more 2018-01-09 15:49:36 pushed 2018-01-09 15:50:28 it's more like actually providing the path than fixing it ;) 2018-01-09 15:50:39 fabled: i also had http://tpaste.us/rZoE 2018-01-09 15:51:13 i just fixed the -other /firmware -> /lib/firmware because it broke my wifi 2018-01-09 16:12:38 ncopa: yes, I'll file it upstream. 2018-01-09 16:23:44 fabled: sorry for the linux-firmware-other breakage :-\ 2018-01-09 16:26:17 https://github.com/alpinelinux/aports/pull/3046 <- I've made this follow-up PR, but I didn't check the IRC yet... I probably need to adjust that linux-firmware-other part again 2018-01-09 16:38:52 fabled: I redid apk help a bit. http://tpaste.us/w7j9 2018-01-09 16:41:07 rebased the PR on master, so fabled's fix is included 2018-01-09 16:43:13 kaniini, looks nice 2018-01-09 16:46:18 making apk --help --verbose work was kinda ugly tho 2018-01-09 16:48:15 oh yes, the ordering can be tricky 2018-01-09 16:48:56 i'm thinking how to do etc/apk/defaults or 'rc' that has the default options 2018-01-09 16:49:27 and possibly that we need to white list options that can be there 2018-01-09 16:49:37 http://tpaste.us/PL0Y 2018-01-09 16:49:44 i think overall the patch series is ok though 2018-01-09 16:51:30 yes, looks pretty good. i was thinking maybe put the group info to struct apk_applet instead 2018-01-09 16:51:31 going to push if no objections 2018-01-09 16:51:46 as for etc/apk/defaults, i don't have good answer for it yet :/ 2018-01-09 16:53:25 would be slightly faster to use an int yes 2018-01-09 16:53:48 ok, i'll do that and then push it 2018-01-09 16:53:51 not about speed. more about adding new applets so you get to specify in applet which group it is. 2018-01-09 16:55:09 sure 2018-01-09 16:59:10 might also make sense to not display some applets in the overview. eg. 'dot' 2018-01-09 17:10:10 dot is helpful for visualizing the dependency tree, but yes, maybe not so great for endusers 2018-01-09 17:10:23 the framework overall is in, we can tweak it a bit :) 2018-01-09 17:19:37 kaniini, thanks 2018-01-09 19:20:03 hi@all 2018-01-09 19:20:33 i got some problems during installing a package 2018-01-09 19:20:38 package mentioned in index not found (try 'apk update') 2018-01-09 19:20:50 but i already tried apk update 2018-01-09 19:21:03 and the package is in apkindex 2018-01-09 19:21:25 any ideas what's wrong or how i can dig deeper into this? 2018-01-09 19:22:55 and of course the apk is also available 2018-01-09 19:23:20 <_ikke_> StarWarsFan: what package? 2018-01-09 19:23:28 its an own build 2018-01-09 19:23:52 building a lot of packages and generating apkindex 2018-01-09 19:23:56 that's working fine 2018-01-09 19:24:18 but some of the packages, which are dependencies of other packages, can't be installed 2018-01-09 19:24:33 not even by hand if i try "apk add " 2018-01-09 19:26:28 are they tagged? 2018-01-09 19:26:40 ? 2018-01-09 19:27:01 in your /etc/apk/repositories something like: 2018-01-09 19:27:02 @jomat /home/jomat/packages/main 2018-01-09 19:27:07 no 2018-01-09 19:27:38 i have this so i have to add my own pkgs with "apk add somepkg@jomat" 2018-01-09 19:28:19 i c 2018-01-09 19:28:26 no, that's not the case here 2018-01-09 19:29:41 no idea what your problem could be… are they found with apk search …? 2018-01-09 19:30:32 I'm also the kind of debugger who fires everything in strace to see what it does, did you try that? :-D 2018-01-09 19:32:12 if not, don't forget -f and -s somehighnumber :-) 2018-01-09 19:35:31 apk search finds the package 2018-01-09 19:36:20 got no knowledge with strace 2018-01-09 19:36:28 :-( 2018-01-09 19:49:44 <_ikke_> apk search relies on the index 2018-01-09 19:49:59 <_ikke_> Are you sure the exact version mentioned in the index is available in the repo? 2018-01-09 19:52:45 yes, i'm sure 2018-01-09 19:53:13 in fact: 2018-01-09 19:53:32 jenkins@a88676587923:/$ apk info cuimenu 2018-01-09 19:53:32 cuimenu-2.0.3-r10 description: 2018-01-09 19:53:32 Eisfair CUI base menu 2018-01-09 19:53:32 cuimenu-2.0.3-r10 webpage: 2018-01-09 19:53:32 http://www.eisfair.org 2018-01-09 19:53:33 cuimenu-2.0.3-r10 installed size: 2018-01-09 19:53:36 917504 2018-01-09 19:53:38 jenkins@a88676587923:/$ sudo apk -U add cuimenu 2018-01-09 19:53:40 fetch http://ftp.halifax.rwth-aachen.de/alpine/v3.7/main/x86_64/APKINDEX.tar.gz 2018-01-09 19:53:42 fetch http://ftp.halifax.rwth-aachen.de/alpine/v3.7/community/x86_64/APKINDEX.tar.gz 2018-01-09 19:53:44 fetch https://web.nettworks.org/e-ng/packages/v3.7/testing/x86_64/APKINDEX.tar.gz 2018-01-09 19:53:47 (1/1) Installing cuimenu (2.0.3-r10) 2018-01-09 19:53:49 ERROR: cuimenu-2.0.3-r10: package mentioned in index not found (try 'apk update') 2018-01-09 19:53:51 1 error; 360 MiB in 145 packages 2018-01-09 19:53:53 jenkins@a88676587923:/$ 2018-01-09 19:54:13 but it's there, see https://web.nettworks.org/e-ng/packages/v3.7/testing/x86_64/ 2018-01-09 19:54:28 https://web.nettworks.org/e-ng/packages/v3.7/testing/x86_64/cuimenu-2.0.3-r10.apk 2018-01-09 20:06:47 <_ikke_> Anyone got expierence building go packages? 2018-01-09 20:07:23 <_ikke_> Trying to update the docker package, but it either complains about dependencies, or it's trapped in a recursive dir loop 2018-01-09 20:07:41 <_ikke_> (the latter mainly because apparently it has a dependency on itself) 2018-01-09 20:36:30 i got it! 2018-01-09 20:36:41 got a wrong arch entry on the dependent package 2018-01-09 20:36:45 <_ikke_> ah 2018-01-09 20:36:45 wrong: 2018-01-09 20:36:49 <_ikke_> that's a nasty one 2018-01-09 20:36:57 arch="noarch" 2018-01-09 20:37:00 right: 2018-01-09 20:37:02 <_ikke_> any 2018-01-09 20:37:05 arch="all" 2018-01-09 20:37:08 <_ikke_> ah, all 2018-01-09 20:37:10 <_ikke_> d'oh 2018-01-09 20:37:30 <_ikke_> Some better feedback would be nice in that case 2018-01-09 20:37:55 ack 2018-01-09 20:50:25 hi, fabled pointed me here, i wanted to see if anyone could implement https://bugs.alpinelinux.org/issues/8307 (it's assigned to him but he's got no time). I'd try to sponsor it because it would make nextcloud on alpine more complete 2018-01-09 20:51:28 <_ikke_> Do you see any challenges in packaging it? 2018-01-09 20:51:46 _ikke_: it's hard to tell for me, i've so far only packaged very simple stuff 2018-01-09 20:52:13 i.e. if it should become a libreoffice subpackage that is probably dependant on how easy it's to do that 2018-01-09 20:52:45 i looked at some of the sources linked in the feature request, but i'm not sure i saw everything 2018-01-09 20:53:20 i think porting the parts in the dockerfiles is survivable, but really uncertain about the office part itself 2018-01-09 20:54:18 <_ikke_> Let me try 2018-01-09 20:54:23 <_ikke_> why are the docker parts necessary? 2018-01-09 20:54:34 they're just reference 2018-01-09 20:54:48 not neccessary 2018-01-09 20:54:59 <_ikke_> ok 2018-01-09 20:56:19 i'm here, /msg me some paypal info whenever you feel like it 2018-01-09 20:57:20 i'm gonna try to update the nextcloud wiki page a bit over the next days. we took turns in setting it up / fixing bugs, so it'll still have errors when i'm done 2018-01-09 20:57:51 <_ikke_> I don't do it for the money 2018-01-09 20:58:29 that's ok but i had already promised it so it would be weird to not give something too 2018-01-09 20:58:40 we can keep that fight for later :) 2018-01-09 20:59:01 i definitely can't package it and it'll fix a large part of our to-be "intranet" 2018-01-09 20:59:22 so i am happy in any possible version 2018-01-09 21:00:21 atm i'm wiping/reflashing my broken phone, a fun thing to do in the evening. 2018-01-09 21:06:47 *wow* the restore seems to work. that's a first. 2018-01-09 21:22:45 <_ikke_> darkfader: hmm, so I see two sets of tags 2018-01-09 21:24:46 what kind of tags? 2018-01-09 21:25:01 sorry maybe i'm in the wrong context right now 2018-01-09 21:25:44 <_ikke_> 2.1.5, 3.0.0, etc and libreoffice-6.0.1 2018-01-09 21:26:08 <_ikke_> I'm looking at https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=libreoffice-online for an example 2018-01-09 21:26:22 <_ikke_> they seem to use the libreoffice-* tags 2018-01-09 21:44:37 it seems the one i pulled in with apk add libreoffice is a 5.4.2 2018-01-09 21:46:13 <_ikke_> It might be that those releases are corresponding with a specific libreoffice release 2018-01-09 21:47:13 yeah especially since they run the same number in there 2018-01-09 21:47:13 pkgver=5.4.2.2 2018-01-09 21:49:29 but if you look here: 2018-01-09 21:49:29 https://aur.archlinux.org/cgit/aur.git/log/PKGBUILD?h=libreoffice-online 2018-01-09 21:49:37 it seems to not update every few weeks 2018-01-09 21:53:19 <_ikke_> Ok, I've got at least ./configure happy for now 2018-01-09 21:54:13 w00t 2018-01-09 22:10:47 <_ikke_> darkfader: building now (after fixing some issues, crossing fingers) 2018-01-09 22:10:56 <_ikke_> oh, next issue 2018-01-09 22:11:48 probably why the arch patch was called "fix annoying errors" :) 2018-01-09 22:12:21 <_ikke_> these have more to do with alpine using musl 2018-01-09 22:12:40 ah, ok 2018-01-09 22:12:59 <_ikke_> It does not know about '__sighandler_t' 2018-01-09 22:20:58 <_ikke_> Ok, fixed that issue, rest for later 2018-01-09 22:22:18 ty again 2018-01-10 06:54:59 good morning at all 2018-01-10 06:57:14 i'm using some docker containers to build packages 2018-01-10 06:57:33 but this is always x86_64 2018-01-10 06:57:51 is it possible to build x86 packages inside of docker containers? 2018-01-10 06:59:55 <_ikke_> StarWarsFan: You need a cross-compiler toolchain 2018-01-10 07:00:12 <_ikke_> Inside our outside of docker doesn't make a difference (a container is just a fancy chroot) 2018-01-10 07:00:35 right but i wouldn't let out these info ;-) 2018-01-10 07:01:49 so is such a toolchain available? i build 32bit packages some years ago using lxc containers on a 64bit host 2018-01-10 07:02:39 but if it would be possible directly without another container inside of the docker-container would be nice 2018-01-10 07:12:55 uh, found this: 2018-01-10 07:12:56 https://hub.docker.com/r/multiarch/alpine/tags/ 2018-01-10 07:13:13 worth a try... 2018-01-10 11:23:27 hi@all 2018-01-10 11:23:41 is there something broken with the edge-git-package? 2018-01-10 11:24:01 is that an actual package? 2018-01-10 11:24:17 trying to setup an edge-buildsystem 2018-01-10 11:24:19 but: 2018-01-10 11:24:29 (38/94) Installing lz4-libs (1.8.0-r1) 2018-01-10 11:24:29 ERROR: git-2.15.0-r1: BAD signature 2018-01-10 11:24:29 (39/94) Installing xz-libs (5.2.3-r1) 2018-01-10 11:24:44 which breaks the whole build 2018-01-10 11:25:09 ah you mean git in edge repo 2018-01-10 11:25:17 are you using cdn? 2018-01-10 11:25:50 in detail building a docker image based on alpine:edge 2018-01-10 11:26:41 around 5hours ago it was working 2018-01-10 11:26:49 are you using cdn? 2018-01-10 11:26:58 what is cdn? 2018-01-10 11:27:10 in repositories file 2018-01-10 11:27:23 i think i know the issue 2018-01-10 11:27:52 don't know, just "FROM alpine:edge" and then my apk install steps... 2018-01-10 11:28:13 looks like recent git commit didnt bump pkgrel 2018-01-10 11:28:17 kaniini: ^ 2018-01-10 11:31:03 ah i c 2018-01-10 11:32:31 the docker image probably uses cdn 2018-01-10 11:32:55 so it fetches a cache of older pkg cause the version didnt change. 2018-01-10 11:33:21 every new build of a package should have a new version or increased pkgrel 2018-01-10 11:33:45 it should be working soon again when pkg is available 2018-01-10 11:34:04 Aerdan: please be aware of ^ 2018-01-10 11:34:25 <_ikke_> clandmeter: This happens quite often it seems 2018-01-10 11:34:48 i hope i dont mix identities 2018-01-10 11:34:59 _ikke_: yes it happens too often. 2018-01-10 11:35:14 cdn users will be affected by it. 2018-01-10 11:36:35 actually this case is different then others. in the past its mostly caused by reverting a package and not increasing pkgrel. 2018-01-10 11:38:54 wasn't my fault. I mean, yes, the person whose fault it was and I both have Japanese names, but that and contributing to alpine are the only things we have in common. :p 2018-01-10 11:39:16 <_ikke_> :D 2018-01-10 11:39:20 ah sorry :) 2018-01-10 11:39:22 <^7heo> Aerdan is japanese? 2018-01-10 11:39:28 nope. 2018-01-10 11:39:30 <^7heo> Ah ok 2018-01-10 11:39:36 identity crisis :) 2018-01-10 11:39:47 ACTION said nothing about ancestry, just names. 2018-01-10 11:40:06 <^7heo> I also asked about the name, not your legacy. 2018-01-10 11:40:29 clandmeter: right, build successful now 2018-01-10 11:40:30 :-) 2018-01-10 11:40:33 thx a lot! 2018-01-10 11:40:42 don't ask ambiguous questions if you want useful answers. dunno what to tell you. :p 2018-01-10 11:40:48 <^7heo> it's spelled "thx a lt" 2018-01-10 11:41:03 <^7heo> Aerdan[m]: your name is Aerdan. I don't see what's ambiguous. 2018-01-10 11:41:39 no, 'Aerdan' is my handle. entirely different. 2018-01-10 16:09:57 ok i bought one of those nvidia jetson tx2's, if I can run regular 32bit arm stuff on it, and not be 29 times slower than x86, it'll do 2018-01-10 16:10:09 i just want compiles to not take a day 2018-01-10 16:11:05 otherwise, maybe I buy a jetson tk1 as a backup 2018-01-10 16:11:26 i'm kind of done using soc's for compiles 2018-01-10 16:13:28 unless anyone has a good 32bit arm suggestion for compiling 2018-01-10 17:39:43 mitchty: we are using xgene 2018-01-10 18:10:53 clandmeter: oof, those look beefy, bit over my budget for home though :) 2018-01-10 18:11:17 i can lie^Wdelude myself into about $800 for stuff, but not that much 2018-01-10 18:48:55 anybody know what that bus error is on armhf? http://build.alpinelinux.org/buildlogs/build-edge-armhf/community/py-lz4/py-lz4-0.17.0-r0.log 2018-01-10 19:15:23 fyi: we are removing linux-hardened in edge 2018-01-10 19:15:28 and upgrading linux-vanilla to 4.14.13 2018-01-10 19:15:29 which have kpti 2018-01-10 19:16:22 right now working on the -vanilla upgrade, but that is our plan for meltdown in edge 2018-01-10 19:16:26 we are still evaluating what to do about stable 2018-01-10 19:16:46 most likely it will be something along the same lines as what we're doing in edge, because the 4.9 patches are junk 2018-01-10 19:18:08 <_ikke_> kaniini: so that means the end for hardened? (at least for now) 2018-01-10 19:18:23 yes 2018-01-10 19:18:27 yes, that is the conclusion we have made 2018-01-10 19:18:57 in the absence of workable patches for hardened and the fact that hardened is vulnerable to meltdown 2018-01-10 19:19:06 we don't think it is a good idea to suggest people continue using it 2018-01-10 19:19:12 <_ikke_> Yeah, makes totally sense 2018-01-10 19:19:43 given the severity of the vulnerability, it would lead people to believe they can depend on something they actually can't depend on (because of meltdown) 2018-01-10 19:20:09 <_ikke_> How about released versions? 2018-01-10 19:20:10 if spender reopens grsecurity or something else comes along we might reintroduce hardened at a later date 2018-01-10 19:20:35 but 2018-01-10 19:20:42 KAISER for 4.9 is not workable and it's not compatible with PaX anyway 2018-01-10 19:21:27 kaniini: 'fix-sync-check.patch': No such file or directory 2018-01-10 19:21:29 do you have that file? 2018-01-10 19:21:31 for released versions, we are still working on it 2018-01-10 19:21:33 yes 2018-01-10 19:21:41 one sec 2018-01-10 19:22:21 dropping (unofficial) grsec is a real downer :-( 2018-01-10 19:22:26 http://tpaste.us/X6ne 2018-01-10 19:22:41 yes, i hope we can bring something back as a replacement for sure 2018-01-10 19:22:41 but we dont have any choice 2018-01-10 19:22:43 ncopa: it's in the name. meltDOWN. 2018-01-10 19:23:05 it's just irresponsible to say "we can depend on hardened" when it has a severe flaw ;) 2018-01-10 19:23:12 yeah, its not a meltUP, thats for sure 2018-01-10 19:24:31 honestly i think it is possible spender drops something just to gloat 2018-01-10 19:25:36 would be nice anyway :) 2018-01-10 19:26:47 Natanael Copa: firmware_install is gone 2018-01-10 19:27:04 hum.. ok... 2018-01-10 19:51:30 ok rebooting into 4.14.13 2018-01-10 19:51:34 brb 2018-01-10 19:54:12 Linux ocelot 4.14.13 #1-Alpine SMP Wed Jan 10 19:27:10 UTC 2018 x86_64 GNU/Linux 2018-01-10 19:54:14 we have a winner 2018-01-10 19:54:38 \o/ 2018-01-10 19:54:59 so looks like it is really KAISER that is utterly broken 2018-01-10 19:54:59 real KPTI seems fine 2018-01-10 19:55:36 did you just comment out the firmware_install? 2018-01-10 19:55:43 or was it renamed 2018-01-10 19:55:50 what happened to it? 2018-01-10 19:56:19 i just removed firmware_install. you're supposed to use linux-firmware package now that is separate 2018-01-10 19:56:22 all firmware was moved there 2018-01-10 19:56:31 bugs : cpu_meltdown 2018-01-10 19:56:32 booooooooo 2018-01-10 19:56:56 need to go buy some AMD thing 2018-01-10 19:57:34 can you test this: https://github.com/dendisuhubdy/meltdown/ 2018-01-10 19:57:40 <_ikke_> 0What's the difference between cpu_meltdown and cpu_insecure? 2018-01-10 19:57:44 <_ikke_> bugs : cpu_insecure 2018-01-10 19:57:55 they renamed it 2018-01-10 19:58:15 <_ikke_> ok 2018-01-10 19:59:14 kaniini: do you have a patch for linux-vanilla? 2018-01-10 19:59:21 i can start look at the 3rd party modules 2018-01-10 20:00:42 git format-patch -1 --stdout | tpaste 2018-01-10 20:00:44 i built it on ppc64le 2018-01-10 20:00:59 would like to test build on aarch64 and armhf too 2018-01-10 20:01:16 Natanael Copa: that github still works, but KPTI wouldnt protect against it anyway 2018-01-10 20:01:29 need retpoline for that one 2018-01-10 20:02:01 omfg git is so much slower 2018-01-10 20:02:03 now 2018-01-10 20:02:18 <_ikke_> it is? :-( 2018-01-10 20:02:35 http://tpaste.us/L6zW 2018-01-10 20:02:53 thanks! 2018-01-10 20:02:56 _ikke_: yes, it is noticeably slower 2018-01-10 20:03:08 git status used to take maybe a second tops 2018-01-10 20:03:10 on aports repo 2018-01-10 20:03:18 now it takes 2 seconds 2018-01-10 20:03:56 :D 2018-01-10 20:04:30 <_ikke_> D: 2018-01-10 20:04:43 ACTION uploaded an image: image.png (41KB)  2018-01-10 20:04:47 this is all i have to say about intel 2018-01-10 20:05:15 <_ikke_> And I have a feeling they are getting away with it too 2018-01-10 20:06:36 not with me they aren't 2018-01-10 20:07:02 <_ikke_> You don't matter for them 2018-01-10 20:07:06 i will not buy any more intel cpus until they prove this bug is fixed in the silicon 2018-01-10 20:07:15 i buy 1000s of intel cpus a year 2018-01-10 20:07:39 do you eat them ? 2018-01-10 20:07:40 there are at least three different class-action lawsuits in the works against Intel over this. 2018-01-10 20:07:40 i might be small fry for them, but if enough mes in this world do the same, it will matter to them 2018-01-10 20:07:56 no, i put them in servers 2018-01-10 20:08:15 kaniini has businesses that require computing. 2018-01-10 20:08:22 what is this flag that disables KPTI 2018-01-10 20:08:31 this is annoying me and i don't care about meltdown 2018-01-10 20:08:33 <_ikke_> Don't recall 2018-01-10 20:08:42 nopti i think 2018-01-10 20:10:09 can also echo 0 to /sys/kernel/debug/x86/pti_enabled 2018-01-10 20:10:15 if you don't want to reboot, that is. 2018-01-10 20:11:09 i mean, i do care about meltdown, but i do not at the same time 2018-01-10 20:11:10 if that makes sense? 2018-01-10 20:11:21 <_ikke_> kaniini: I totally feel you 2018-01-10 20:11:50 for each machine you need to decide what is more important: being speedy or preventing meltdown 2018-01-10 20:12:00 and then set the kernel flag 2018-01-10 20:12:04 decisions, decisions 2018-01-10 20:13:10 how about i just turn it on when i enter my CC number 2018-01-10 20:13:12 hi@all 2018-01-10 20:13:13 ;) 2018-01-10 20:13:20 <_ikke_> heh 2018-01-10 20:13:25 may i ask a question regarding a strange build issue? 2018-01-10 20:14:07 hi StarWarsFan. ofc! 2018-01-10 20:14:17 dont ask i you can ask. just ask! :) 2018-01-10 20:14:23 if* 2018-01-10 20:14:28 i setup some docker containers to build packages 2018-01-10 20:15:02 and there are some boxes with edge-x86, v3.7-x86 and v3.6-x86 2018-01-10 20:15:13 based on the images from here: 2018-01-10 20:15:19 https://hub.docker.com/r/multiarch/alpine/tags/ 2018-01-10 20:15:43 no i have some builds, which hangs on the configure step 2018-01-10 20:15:47 but only on the v3.7 box 2018-01-10 20:15:52 on edge it works 2018-01-10 20:15:55 and on 3.6 too 2018-01-10 20:16:00 upgrade bash 2018-01-10 20:16:06 see https://web.nettworks.org/ci/job/e-ng/job/stable/job/testing/job/x86/job/greyfix/3/console 2018-01-10 20:16:10 v3.7 bash was broken 2018-01-10 20:16:34 it is fixed now 2018-01-10 20:16:42 ah, interesting! 2018-01-10 20:16:50 thx for the hint, will try 2018-01-10 20:17:56 i think rebuilding the image should give you newer bash? 2018-01-10 20:18:05 don't look at me, i haven't any clue how docker works ;) 2018-01-10 20:18:21 im a bit surprised it still is an issue 2018-01-10 20:18:30 i thought we fixed it long time ago 2018-01-10 20:19:39 on x86_64 image it works 2018-01-10 20:19:51 but i got the problem on x86 image 2018-01-10 20:20:02 i386-v3.7 2018-01-10 20:22:18 kaniini: to which version? 2018-01-10 20:22:25 actually it looks like this: 2018-01-10 20:22:26 <_ikke_> kaniini: docker is just a fancy chroot :P 2018-01-10 20:22:28 jenkins@3d2a0c4df10f:/$ apk info bash 2018-01-10 20:22:28 bash-4.4.12-r2 description: 2018-01-10 20:22:28 The GNU Bourne Again shell 2018-01-10 20:22:28 https://www.gnu.org/software/bash/bash.html 2018-01-10 20:22:28 bash-4.4.12-r2 webpage: 2018-01-10 20:22:30 bash-4.4.12-r2 installed size: 2018-01-10 20:22:32 1490944 2018-01-10 20:32:57 hm, there's nothing to update 2018-01-10 20:33:13 weird 2018-01-10 20:33:41 what version do you have? 2018-01-10 20:33:42 apk version bash 2018-01-10 20:34:02 should be at -r5 2018-01-10 20:34:10 (on edge) 2018-01-10 20:34:10 Natanael Copa: how is it coming along with the modules ? 2018-01-10 20:34:11 i'm working on doing the -virt profile 2018-01-10 20:34:16 jenkins@3d2a0c4df10f:/$ apk version bash 2018-01-10 20:34:16 bash-4.4.12-r2 = 4.4.12-r2 2018-01-10 20:34:16 Installed: Available: 2018-01-10 20:35:41 btw, here's aonther one that hangs: https://web.nettworks.org/ci/job/e-ng/job/stable/job/testing/job/x86/job/libite/5/console 2018-01-10 20:38:39 on edge it works: https://web.nettworks.org/ci/job/e-ng/job/edge/job/testing/job/x86/job/libite/lastBuild/console 2018-01-10 20:39:14 and on 3.6 too: https://web.nettworks.org/ci/job/e-ng/job/previous/job/testing/job/x86/job/libite/lastBuild/console 2018-01-10 20:39:25 the jobs fix patch may be required then 2018-01-10 20:39:47 ? 2018-01-10 20:41:45 let me look 2018-01-10 20:42:18 no 2018-01-10 20:42:19 its there 2018-01-10 20:42:22 in -r2 2018-01-10 20:44:12 I mean https://git.alpinelinux.org/cgit/aports/tree/main/bash/fix-jobs.patch 2018-01-10 20:46:09 so it might be another problem... 2018-01-10 20:48:03 yes, it is there in -r2 2018-01-10 20:51:35 and now? other ideas? 2018-01-10 20:52:50 you might try strace it 2018-01-10 20:54:05 checking whether modules can be built... no 2018-01-10 20:54:06 configure: error: *** Unable to build an empty module. 2018-01-10 20:54:08 need figure out why 2018-01-10 20:55:37 Makefile:942: *** "Cannot generate ORC metadata for CONFIG_UNWINDER_ORC=y, please install libelf-dev, libelf-devel or elfutils-libelf-devel". Stop. 2018-01-10 20:55:50 so what do you need? what should i do now? 2018-01-10 20:56:57 no, per the log that patch is -r5 2018-01-10 20:58:17 Aerdan[m]: it's in -r2 on 3.7-stable 2018-01-10 20:58:33 Aerdan[m]: we're talking about 3.7, not edge 2018-01-10 21:00:14 Natanael Copa: libelf-dev probably needs added to linux-vanilla-dev 2018-01-10 21:00:44 i already did 2018-01-10 21:00:47 elfutils-dev 2018-01-10 21:01:12 i'll test it and add it to your commit 2018-01-10 21:03:22 probably put 2018-01-10 21:03:40 provides="linux-hardened=$pkgver linux-grsec=$pkgver" to force the upgrade 2018-01-10 21:03:55 need to do that to the modules too 2018-01-10 21:03:56 bah 2018-01-10 21:08:07 ? 2018-01-10 21:15:37 StarWarsFan: you might try to strace it, like i said 2018-01-10 21:19:48 right and my question was, how to go ahead. i have no idea what to do now 2018-01-10 21:20:06 run configure step with strace? how? 2018-01-10 21:20:21 strace ./configure 2018-01-10 21:20:32 ok, thx 2018-01-10 21:20:34 just be warned 2018-01-10 21:20:38 you will get LOTS OF OUTPUT 2018-01-10 21:25:11 strace: ptrace(PTRACE_TRACEME, ...): Operation not permitted 2018-01-10 21:26:58 StarWarsFan: hardened kernel? 2018-01-10 21:28:31 i have no idea, it's a container based on https://hub.docker.com/r/multiarch/alpine/tags/ 2018-01-10 21:28:40 no i mean for the host system 2018-01-10 21:28:42 i386-v3.7 2018-01-10 21:29:21 host is a debian 9 vm running on proxmox 2018-01-10 21:31:15 Linux flap 4.9.0-4-amd64 #1 SMP Debian 4.9.65-3+deb9u1 (2017-12-23) x86_64 GNU/Linux 2018-01-10 21:39:11 sry, very late, need some sleep 2018-01-10 21:39:18 will read backlog 2018-01-10 21:55:12 wow. all 3rd party modules built 2018-01-10 21:57:47 ok i think i will push kernel 4.14 2018-01-10 21:58:29 kaniini: ok to push your 4.14 upgrade? 2018-01-10 21:58:33 with 3rd party modules 2018-01-10 21:58:36 it boots in qemu 2018-01-10 22:39:32 Natanael Copa: eys, i am okay with it 2018-01-10 22:40:50 im just waiting for armhf build to complete 2018-01-10 22:41:06 i suppose i can push it, the other arches built 2018-01-10 22:41:48 pushed 2018-01-10 22:54:04 ncopa: i can try it on a few boxes, which repo mirror would i need to see it? 2018-01-10 22:54:33 i pushed it to edge 2018-01-10 22:55:05 mmh, then i'll just use some test vms :) 2018-01-10 23:00:27 one weird thing i managed to produce twice is that it adds a lot of fw packages 2018-01-10 23:00:30 https://hastebin.com/xalidoxaqu.cs 2018-01-10 23:00:39 not related to the kernel, just weird 2018-01-10 23:01:35 ah and i need one more update to really run edge, give me a few minutes :) 2018-01-10 23:03:54 ncopa: no luck, i'm on http://nl.alpinelinux.org/alpine/edge/main 2018-01-10 23:04:02 and Linux alpine-base 4.9.73-0-hardened #1-Alpine SMP Tue Jan 2 16:33:46 UTC 2018 x86_64 GNU/Linux 2018-01-10 23:04:54 probably not up to date yet, i'll check again tomorrow and let you know if i can break anything 2018-01-10 23:27:23 darkfader: the hardened kernel will likely not be updated 2018-01-10 23:27:33 we only update the vanilla kernel 2018-01-10 23:27:45 and will probably remove the hardened kernel :-( 2018-01-10 23:29:42 ? 2018-01-10 23:30:00 i wasn't aware the hardened kernel was being phased out 2018-01-10 23:30:30 well, it is. thank Intel for being incompetent chip designers. 2018-01-10 23:30:37 o 2018-01-10 23:30:40 ofc 2018-01-10 23:30:40 wew 2018-01-10 23:31:36 and the people/companies who drove grsecurity off the grid 2018-01-10 23:32:29 ncopa: ah ok 2018-01-10 23:32:32 shit 2018-01-10 23:32:33 :> 2018-01-10 23:33:16 oss pricing is a funny thing. one of my former bosses is getting really rich because we came up with a good pricing model 2018-01-10 23:33:29 it is affordable 2018-01-10 23:33:44 this... ah well. nothing one can do 2018-01-10 23:34:57 this meltdown issue is exactly why we need grsecurity 2018-01-10 23:35:25 the fixes the push for it is rushed and seems to have various vulnerabilities 2018-01-10 23:35:47 23:34 <@ncopa> this meltdown issue is exactly why we need grsecurity 2018-01-10 23:35:47 23:35 <@ncopa> the fixes the push for it is rushed and seems to have various vulnerabilities 2018-01-10 23:35:56 i can't even copy with my flu 2018-01-10 23:36:08 was gonna send it to a friend :) 2018-01-10 23:36:40 i agree so much, it's much easier if there's a few working layers of security 2018-01-10 23:37:18 its not realistic to make the kernel bug free, so we need things like grescurity that makes it really hard (or impossible) exploit the bugs when they appear 2018-01-10 23:38:36 and adjustable, so that if 64kb distance isn't enough, make it 1mb... which is just the thing one could do for the last disaster 2018-01-10 23:38:38 reading the grsecurity twitter doesnt really fill me with hope 2018-01-10 23:38:58 they honestly sound quite close to simply losing it entirely over something tiny and becoming a goat farmer 2018-01-10 23:39:41 if you invented houses that don't burn and everyone tells you so what, we put water in our basement fuck off 2018-01-10 23:39:59 and then after 10 years of fires they go fuck that! we put it in the walls too! 2018-01-10 23:40:04 i'd go nuts too 2018-01-10 23:40:18 sticking to it is just scf 2018-01-10 23:40:37 Arcaire: unless you don't see this as a LKML issue 2018-01-10 23:40:51 i just see that dude as particularly screechy tbh 2018-01-10 23:40:52 thats all 2018-01-10 23:40:55 but a worldwide fuckup that damages the whole world 2018-01-10 23:40:58 which is the truth 2018-01-10 23:41:11 "something tiny" that's annoying can be pretty big after a decade 2018-01-10 23:41:40 except its pretty much always been like that as far back as i can go, it seems 2018-01-10 23:41:47 Arcaire: i agree with you, reading twitter is not very encouraging 2018-01-10 23:41:53 mega self-righteous and screechy and then wonders why the fuck nobody wants to work with him 2018-01-10 23:42:21 probably because it's mostly volunteer and people don't wanna be around that sort of personality idk it makes sense 2018-01-10 23:42:27 same reason nobody speaks to freebsd developers 2018-01-10 23:42:29 they're also awful 2018-01-10 23:42:35 Arcaire: like uh, wait, intel 2018-01-10 23:42:38 no 2018-01-10 23:42:40 not like intel at all 2018-01-10 23:42:45 who is no longer wishing to work with GPL folks? 2018-01-10 23:42:49 and comes to bsd confs 2018-01-10 23:43:04 just fyi 2018-01-10 23:43:41 and the volunteer card is so old these days as people build careers on their OSS work. without being responsible 2018-01-10 23:43:44 i can't really muster up the energy to care about the point you're trying to make tbh 2018-01-10 23:43:45 sorry 2018-01-10 23:43:52 ok 2018-01-10 23:43:56 grsecurity's reputation is self-inflicted 2018-01-10 23:44:02 ^ 2018-01-10 23:44:03 It's not like refusing to interact with the open-source community is anything new. see: nVidia, ARM, Qualcomm, ... 2018-01-10 23:44:54 yeah but i find it sick that the low *interest* in security of the kernel people is waived away with the "they're volunteers" flag 2018-01-10 23:45:01 they go to work, they get paid for it 2018-01-10 23:45:04 no they don't 2018-01-10 23:45:06 source: me 2018-01-10 23:45:08 if spender had been nicer to the community and had a positive attitude this would be a wonderful opportunity for him to promote his work 2018-01-10 23:45:14 i go to work and i get paid to do security shit on bespoke systems 2018-01-10 23:45:17 i don't want to deal with people like that guy 2018-01-10 23:45:19 /ever/ 2018-01-10 23:45:21 and in fact, i don't 2018-01-10 23:45:25 but he's been on a negative trend for years 2018-01-10 23:45:25 if clients act like that, they get fired 2018-01-10 23:45:50 employees, too. 2018-01-10 23:45:52 instead, he's witholding regressions in KAISER/KPTI out of spite, just like he regularly does 2018-01-10 23:46:04 like damn son you really gonna act like that right now? 2018-01-10 23:46:08 and instead of "ok, this is exactly why you should use grsecurity, come here guys I have the solution for you", he's going full "I told you so, fuck you" 2018-01-10 23:46:15 so obviously that's not exactly working 2018-01-10 23:46:19 yeah 2018-01-10 23:46:56 i don't mean we need him to fix that. but i think there's a major issue when linus says he can't stop and think about implications of security issues or he'd not get anything done 2018-01-10 23:47:21 tbh, grsec, and infosec in general, seems to be all about ego, for him. 2018-01-10 23:47:24 if "ze upstream" goes 5% to crazy spenders direction, we'll be fine 2018-01-10 23:47:28 which is really not somethign you want in a security guy. 2018-01-10 23:47:36 yeah :( 2018-01-10 23:47:53 Aerdan[m]: actually that's a really good way to put it 2018-01-10 23:47:58 thanks, that's the terms i've been searching for 2018-01-10 23:48:10 that's why i think the non-security people need to up their game so it's not some one-man security gurus to try to fix it 2018-01-10 23:48:54 i find it horrific that one person can consistently run circles around a large team, as it is. 2018-01-10 23:50:27 honestly I'd be a lot happer if he'd spent his time making his patches mainline-able. 2018-01-10 23:53:26 i think as long as it depends on someone like him, we got a problem 2018-01-10 23:53:52 and not because of his personality, but because it means the average kernel code is just not getting enough oversight 2018-01-10 23:54:44 and i also think this is why he's off balance, or whatever you may call it 2018-01-11 00:05:19 i am rebooting to 4.14. brb hopefully 2018-01-11 00:05:26 lolgl 2018-01-11 00:08:41 Linux ncopa-macbook 4.14.13 #1-Alpine SMP Wed Jan 10 22:42:04 UTC 2018 x86_64 Linux 2018-01-11 00:08:45 boots 2018-01-11 03:22:17 Question: is it a future roadmap feature or such for APK to support multiple packages that provide the same name but different versions? 2018-01-11 03:22:46 For instance, if python2 provides="python=2.7" and python3 provides="python=3.6". Or Ruby, or LLVM, etc 2018-01-11 03:32:00 A. Wilcox: we have thought about it (and technically the solver itself already supports this), maybe after 2.10. patches would be interesting. 2018-01-11 03:35:26 My specific thought is slotting kernels so that eventually modules could be 'left behind' for the currently installed kernel 2018-01-11 03:35:55 I've just changed easy-kernel packaging in Adélie to have the package name "easy-kernel-$pkgver-$pkgrel" and provides="easy-kernel=$pkgver-r$pkgrel" 2018-01-11 03:36:28 do you mean firmware? 2018-01-11 03:36:30 And I realised after I did that and sent it to the builders that APK probably doesn't support that yet 2018-01-11 03:37:31 Aerdan, no. Consider: `apk upgrade`, kernel and modules are upgraded with the other packages, you don't have time to reboot immediately. Now you plug your cell phone in to use for tethering for something important. 2018-01-11 03:37:43 Oops, usb_cdc isn't available. 2018-01-11 03:37:47 Sorry, no tether for you 2018-01-11 03:38:11 This is surprisingly common. Two people have already hit it in Adélie and we only have 25 users. I've also seen it multiple times on Alpine #support 2018-01-11 03:38:52 A. Wilcox: yes we want a solution for that in apk 2018-01-11 03:39:45 by the way 2018-01-11 03:39:47 I'll see what I can cook up, then. 2018-01-11 03:40:13 need people to test linux-vanilla 4.14.13 2018-01-11 03:40:24 because next step is to transition linux-hardened to it 2018-01-11 03:48:42 anything you want in particular? 2018-01-11 03:48:48 i run it on all of my boxen 2018-01-11 03:48:53 happy to push to edge for that if you want 2018-01-11 04:05:51 yes, just make sure it works for you 2018-01-11 04:06:22 we are upgrading all kernels to linux-vanilla 4.14.13, this is what the core team has decided to do about meltdown/spectre 2018-01-11 05:16:17 my alpine docker builds are failing right now, because http://dl-cdn.alpinelinux.org/alpine/v3.7/main/x86_64/APKINDEX.tar.gz is not working 2018-01-11 05:16:42 same for http://dl-cdn.alpinelinux.org/alpine/v3.7/community/x86_64/APKINDEX.tar.gz 2018-01-11 05:16:47 they redirect to http://110.164.252.137/wpwarn/soft/wpcbt_res.php 2018-01-11 05:27:59 why is that IP a result in google search for "wpcbt_res.php" 2018-01-11 05:28:00 wth 2018-01-11 05:29:38 kaniini: 2018-01-11 06:06:36 don't look at me, i keep wanting to replace fastly with my own anycast stuff :) 2018-01-11 07:34:39 Is there a way to tell abuild to go away with its auto deps? 2018-01-11 07:35:13 I finally fixed our easy-kernel-modules package's symlink for /lib/modules/$(uname -r)/source, and now abuild thinks easy-kernel-modules should install easy-kernel-src 2018-01-11 07:35:22 That's not a good thing 2018-01-11 10:02:01 is it just my box or clang in AL 3.6 is slightly broken? `clang -dM -E - (in AL 3.7 there is no problem) 2018-01-11 12:36:00 kaniini: AWilcox[m]: I’d like to set up some systematic convention for packaging multiple versions of one project (e.g. llvm, python, …); currently we add version suffix to the version, e.g. llvm3.9, imagemagick6, …, but it’s not possible to reliable determine if numeric suffix is part of the name or a version suffix; I suggest @version, e.g. llvm@3.9, imagemagick@6 2018-01-11 12:36:37 kaniini: AWilcox[m]: adding provides="llvm=3.9" 2018-01-11 12:37:50 kaniini: AWilcox[m]: and also force convention that version suffix is used only for older versions, not last, currently it’s chaotic 2018-01-11 12:38:18 kaniini: AWilcox[m]: the current state is problematic e.g. for checking new upstream versions 2018-01-11 12:46:52 the only way to fully support multiple version packaging is to install software in its own directory. 2018-01-11 12:47:08 AWilcox[m]: " Is there a way to tell abuild to go away with its auto deps?" … try `options="!tracedeps"`, but I’m not sure if it’s what you want 2018-01-11 12:47:10 numbering is a thing, but the real issue is, what to do with executables? 2018-01-11 12:47:23 what should /usr/bin/llvm point to? 2018-01-11 12:47:30 what version should it be? 2018-01-11 12:47:46 skarnet: we already handle it 2018-01-11 12:47:52 oh? how? 2018-01-11 12:47:59 skarnet: look into llvm packages for exsample 2018-01-11 12:48:27 uh, not going to go sleuthing into the llvm APKBUILD 2018-01-11 12:48:32 skarnet: I’m on the meeting right now, so don’t have time to discuss it 2018-01-11 12:48:39 ok. 2018-01-11 18:08:14 question on the multi version stuff 2018-01-11 18:08:38 as i'd be impacted and also have a recently relevant and impacted use of it 2018-01-11 18:08:58 ok so turns out the arm problem for getting ghc built on armhf is a bug 2018-01-11 18:09:23 so what I'd like to be able to do is say: for x86 ghc 8.2.2 is kosher, for armhf stick to 8.02 for now 2018-01-11 18:09:39 with both providing some meta ghc package 2018-01-11 18:09:54 just another case where it can get messay :) 2018-01-11 18:34:12 jirutka: @version won't work because @ is used for tags 2018-01-11 18:34:28 kaniini: ah, right 2018-01-11 18:34:32 kaniini: that’s pitty :/ 2018-01-11 18:34:51 you could use a different sigil 2018-01-11 18:36:36 any suggestions? 2018-01-11 18:37:16 % maybe? 2018-01-11 18:37:46 # won't work unless escaped, or I'd suggest that. 2018-01-11 18:38:11 dunno. on adelie side of things we're in a holding pattern right now because awilfox is having a family emergency 2018-01-11 18:50:01 Hi just tested the updated vanilla kernel, before this always used "virthardened" noticed a difference in kernel local version. In the "hardended" kernels $pkgrel is always visibile in the version string but not for vanilla. Intentional? 2018-01-11 18:51:53 4.9.73-0-hardened vs 4.14.13 but "should be" 4.14.13-$pkgrel i.e. -0 or -1 ... 2018-01-11 18:52:46 this is causing problems with scripts checking if a reboot is needed due to that a kernel with higher pkgrel was loaded 2018-01-11 18:55:46 HRio: that's actually kABI 2018-01-11 18:55:52 HRio: but 2018-01-11 18:56:02 HRio: we are going to add all of that stuff :) 2018-01-11 18:57:30 echo "-$pkgrel-$i" > "$srcdir"/build-$i/localversion-alpine \ in linux-hardened/APKBUILD 2018-01-11 18:57:42 yes, aware 2018-01-11 19:00:51 we are just looking to verify that KPTI does not cause systems to hang randomly 2018-01-11 19:00:53 because KAISER does 2018-01-11 19:00:53 OK, good, will wait a bit with updating the scripts until things settle a bit, then? 2018-01-11 19:00:54 :D 2018-01-11 19:01:17 OK boots as XEN 64-bit PV guest at least ;-) 2018-01-11 19:03:15 This is for the apt-dater-host package, one of the functions only handle "flavour" kernels it seems, as that is what I have been running all along 2018-01-11 19:06:20 should there not be any nr_overhead counter in /proc/vmstat for KPTI? only KAISER? 2018-01-11 19:08:39 no idea 2018-01-11 19:08:59 KPTI has cpu_meltdown in /proc/cpuinfo 2018-01-11 19:09:35 https://patchwork.kernel.org/patch/10143301/ 2018-01-11 19:24:12 KPTI does not do that 2018-01-11 19:25:39 ah, thanks. The 4.14 kernel runs fine on my chromebook pixel as well. Thanks! 2018-01-11 19:30:03 I'm going to have to seriously consider migrating one of my servers off of ubuntu 16.04 LTS, because they're still sitting on 4.4 2018-01-11 19:34:56 HRio: -virt profile coming shortly. working on it :) 2018-01-11 19:41:36 thanks, only have XEN dom0:s and my laptop that are non -virt. The rest are XEN domUs or vmware guests. 2018-01-11 20:05:10 raccoon:~$ uname -a 2018-01-11 20:05:11 Linux raccoon 4.14.13-1-vanilla #2-Alpine SMP Thu Jan 11 19:36:27 UTC 2018 x86_64 GNU/Linux 2018-01-11 20:05:19 is this better for you? :) 2018-01-11 20:08:40 shouldn't that be r$pkgrel-$flavor? 2018-01-11 20:08:48 needs more r 2018-01-11 20:11:02 but /w 26 2018-01-11 20:12:07 Shiz: it matches the hardened one 2018-01-11 20:12:20 Shiz: he indicated he needs output to be same 2018-01-11 20:12:31 something something hardened should be fixed too then 2018-01-11 20:12:33 ;p 2018-01-11 20:12:46 Shiz: hardened is being fixed with rm -rf; see alpine-devel 2018-01-11 20:50:39 kaniini: sorry AFK: 4.4.59-0-virtgrsec, 4.9.73-0-hardened etc. is handled fine by the script so 4.14.13-1-vanilla will be perfect 2018-01-11 20:52:25 what to do about the image names? all flavour kernels are called vmlinuz- initramfs- but vanilla does not follow this pattern. 2018-01-11 20:53:05 for vanilla vmlinuz / initramfs-vanilla, should it not be vmlinuz-vanilla? 2018-01-11 20:56:09 it is vmlinuz-vanilla now 2018-01-11 20:56:25 -rw-r--r-- 1 root root 4752432 Jan 11 19:54 vmlinuz-vanilla 2018-01-11 20:56:26 :) 2018-01-11 20:58:12 really nice, thanks 2018-01-11 21:04:49 can test boot the -virt flavour when available as guest on vmware esxi 6.5 and as domU on XEN 4.4, 4,7 & 4.8. If you are not testing that already. 2018-01-11 21:05:10 working on the modules atm 2018-01-11 21:11:14 OK, I'm off to bed now. Again big thanks for your work on 4.14, much appreciated! 2018-01-11 21:12:32 this is part of an overall redesign of the kernel packaging 2018-01-11 21:52:24 kaniini: you don’t have configured git hooks for aports (https://github.com/alpinelinux/aports#git-hooks), do you? ;) 2018-01-11 21:55:14 seems like the thing to mention on alpine-devel 2018-01-11 21:55:28 i do now :) 2018-01-11 21:55:47 kaniini: I hoped that people will recognize it in so short README… 2018-01-11 21:56:01 i've never read the README :D 2018-01-11 21:56:51 and even didn’t notice new top-level directory? :P 2018-01-11 21:57:21 nope 2018-01-11 21:57:29 my interactions with the aports git repo mostly involve doing whatever i need to do right now 2018-01-11 21:57:44 i suspect it is the same for most people 2018-01-11 21:58:46 well, with any luck, there will be a linux-virt in the repos in about 10 minutes 2018-01-11 21:59:29 hm, my experience is maybe different b/c I watch GitHub feed more often than Twitter 2018-01-11 22:19:17 my github feed is entirely unusable 2018-01-11 22:24:40 I’m very selective about what project I follow 2018-01-11 22:25:19 yes, there is this one called alpinelinux/aports 2018-01-11 22:25:21 every time i look at my feed 2018-01-11 22:25:28 900+ notifications 2018-01-11 22:29:25 I've given up "watching" that repo after one day :p 2018-01-11 22:34:01 anyone with clang on AL 3.6 within reach? would like to confirm whether `clang -dM -E - I never liked dichotomy regarding special treating of vanilla vs grsec/hardened. thankfully rest is generic enough, so -vanilla suffix should be rather painful as long as update-extlinux (or equiv on non-x86 archs) is called after upgrade 2018-01-11 22:51:09 przemoc: negative 2018-01-11 22:51:29 Shiz: regarding clang? 2018-01-11 22:51:32 przemoc: i could use some assistance with getting the hardened-only modules built on vanilla if you have time 2018-01-11 22:51:34 yes 2018-01-11 22:51:54 damn, I feared that. thanks, Shiz 2018-01-11 22:52:17 Shiz: you're sure it's AL 3.6? because I know that 3.7 is fine 2018-01-11 22:52:19 oh 2018-01-11 22:52:23 fug, no you're right 2018-01-11 22:52:25 that's 3.7 2018-01-11 22:52:27 :P 2018-01-11 22:58:42 kaniini: not sure if sevaral minutes will be enough, but I guess I could help you (if I'm knowledgable enough) as long as testing on real hw is not needed, because I don't have any real edge instance. 2018-01-11 22:59:36 (only lxc one on x86-64) 2018-01-11 23:01:46 by the way i am looking at strcat's linux-hardened as an alternative 2018-01-11 23:01:52 but for now we drop grsecurity since it's vulnerable and we can't fix it ourself 2018-01-11 23:07:11 unfortunately I have no knowledge about existing hardening efforts and which seems to be the best to switch to. in the past I only very loosely followed grsec/PaX efforts, and rather from user's PoV than programmer's, so I'm afraid I'm not a good person for providing valuable input regarding that matter. 2018-01-11 23:07:51 Shiz seemed to be more active regarding hardening matters IIRC, so he may be a better person here for that 2018-01-12 05:59:49 remaining modules: dahdi-linux-hardened, devicemaster-linux-hardened, virtualbox-guest-modules-hardened, virtualbox-additions-hardened, sch-cake-hardened, ipt-netflow-hardened 2018-01-12 06:07:39 kaniini: tested the linux-virt 4.14.13-2-virt kernel on Alpine 3.7 guests this morning. Under XEN 4.8 and VMware 6.5 the VMs boots just fine. 2018-01-12 07:18:22 VOISE 2018, the battle continues!!!!!!!!!!!!!!!!!!!!! 2018-01-12 07:18:29 livecoin.net 2018-01-12 07:18:35 best exchange, great service, intuitive interface, trustworthy, secure 2018-01-12 07:18:39 go BTC to VOISE 2018-01-12 07:18:47 As the drillers don't stop drilling until they strike black gold 2018-01-12 07:18:56 in the blood soaked streetz of America 2018-01-12 07:18:58 America 2018-01-12 07:19:06 the land of the free 2018-01-12 07:19:14 in America 2018-01-12 07:19:17 the niggaz don't stop 2018-01-12 07:19:20 in America 2018-01-12 07:19:24 America 2018-01-12 07:19:39 I'm out with the Boozie Badazz - Wake Up! 2018-01-12 07:19:45 my niggaz! 2018-01-12 07:19:48 my dorks! 2018-01-12 07:20:04 my savage brothaz! 2018-01-12 08:15:56 <_ikke_> ugh 2018-01-12 09:06:46 win 3 2018-01-12 09:47:02 przemoc: around? 2018-01-12 11:40:04 hello 2018-01-12 12:10:51 heyo 2018-01-12 12:56:39 hi 2018-01-12 12:56:41 ACTION sent a long message: ncopa_2018-01-12_12:56:40.txt  2018-01-12 12:57:15 i think this is the fix? http://tpaste.us/05LK 2018-01-12 14:59:23 Natanael Copa: no, removing the provides is the fix for now 2018-01-12 15:04:29 kaniini: something like this? 2018-01-12 15:04:29 http://tpaste.us/qoKO 2018-01-12 15:10:32 thats what i would do yes 2018-01-12 15:10:40 until we are ready to flip the switch 2018-01-12 15:10:49 because some modules arent ported yet 2018-01-12 15:11:54 *nod* 2018-01-12 15:37:51 we should update linux-headers package too 2018-01-12 17:30:47 i dunno why devicemaster-linux-vanilla fails to install makedepends 2018-01-12 17:38:21 it fails to resolve linux-vanilla-dev apparently 2018-01-12 17:38:46 linux-vanilla-dev=4.14.13-r0 was fine 2018-01-12 18:42:14 Is anyone building a newer 32 bit OpenSUSE? 2018-01-12 18:49:28 I wonder how hard it would be to modifie openSUSE 42.2. 2018-01-12 19:19:11 KingParrot: the issue with 32-bit is that people are linking to the single-.so version of LLVM rather than to the LLVM components they actually need, and the single-.so version exhausts elf32 addressable space. 2018-01-12 19:19:49 intresting. 2018-01-12 19:19:59 and unfortunately you need LLVM to do X because GLX is required, and Mesa requires LLVM in order to (pre)compile GLSL shit. 2018-01-12 19:20:05 how is LLVM any diffrent then LLLVM? 2018-01-12 19:21:00 opps i ment LLVM - LLVM they look the same to me. 2018-01-12 19:21:46 maybe I should learn much about what is behind compliling. 2018-01-12 19:22:44 (whether that exhaustion can be resolved or not depends on whether LLVM can be forced to not build with debugging symbols.) 2018-01-12 19:23:56 So it could be very crafty to get such an event to start such a proccess? 2018-01-12 19:25:14 ...what? 2018-01-12 19:28:02 Natanael Copa: -vanilla has a tagged kABI like hardened now, as part of the -virt profile 2018-01-12 19:28:54 Aerdan do u work with python? 2018-01-12 19:31:50 I don't know what that has to do with LLVM, but yes...? 2018-01-12 19:38:01 I was trying to install python3-tk but it would not instal do to depandancy issues. 2018-01-12 19:40:53 okay, what do you need tkinter for? 2018-01-12 19:41:37 I want to attempt to make my own GUI 2018-01-12 19:42:24 KingParrot: what does apk say when you get the dependency issues. 2018-01-12 19:43:00 I would strongly recommend python3-qtpy over python3-tk if you're looking to write GUI applications. 2018-01-12 19:43:19 py3-qtpy* 2018-01-12 19:43:19 Aerdan: that's not helpful, his apk is somehow broken 2018-01-12 19:43:46 yes, that part I'm leaving up to you. 2018-01-12 19:44:18 ok I wil attempt to install it because I don't recall exactly 2018-01-12 19:44:39 was it an "unsatisfiable constraints" error? 2018-01-12 19:44:43 if so, that means the package name you put in is wrong 2018-01-12 19:44:57 oh it is in my test notes 2018-01-12 19:45:06 no need to make a new attempt 2018-01-12 19:45:58 it may be a good idea to change the unsatisfiable constraints error when trying to add unknown package names 2018-01-12 19:46:09 What is the diffrance of py3-qtpy vs pyqt? 2018-01-12 19:47:47 Peppermint 8 testblt (>= 2.4z-9)libtcl8.6 (>= 8.6.0)tk8.6-blt2.5 (>= 2.5.3) < Many search and could not find. 2018-01-12 19:48:01 it is not the most obvious error message 2018-01-12 19:48:15 qtpy gives you a consistent API between PyQt4, PyQt5, and PySide. 2018-01-12 19:48:23 sigh 2018-01-12 19:49:00 I was told to use pip 2018-01-12 19:49:07 KingParrot: what was in /etc/apk/repositories 2018-01-12 19:50:36 just `apk add python3-tkinter` 2018-01-12 19:50:50 it works for me 2018-01-12 19:51:09 (vultr's LA DC needs to get their shit together...) 2018-01-12 19:53:39 I don't have a apk dir 2018-01-12 19:55:14 apk is a command not a directory. 2018-01-12 20:18:33 Python 2.4 Fedora Core 3 users can download RPMs, or build from source. 2018-01-12 20:29:36 alpine does not use RPMs 2018-01-12 20:34:41 looks like I found python-wxgtk2.8 as a deb now the question is will it install and work 2018-01-12 20:34:58 does alpine use rpm? 2018-01-12 20:35:15 Does alpine run live off a usb? 2018-01-12 20:35:32 opps I meant to ask does alpine use deb 2018-01-12 20:36:02 seems that most distros are rpm or deb 2018-01-12 20:37:19 alpine uses apk, which is basically a redesign of the dpkg/apt package management stack combined with an archlinux-like build system called APKBUILD 2018-01-12 20:39:46 oh intresting 2018-01-12 20:40:09 I never herd of apk before 2018-01-12 20:40:20 I have herd of PACMAN 2018-01-12 20:40:34 so apk is a file system then? 2018-01-12 20:40:51 opps I ment apk is a package manager then? 2018-01-12 20:41:22 https://wiki.alpinelinux.org/wiki/Alpine_Linux_package_management 2018-01-12 20:41:33 k 2018-01-12 20:43:10 #alpine-linux may be a better channel for your inquiries 2018-01-12 20:44:33 k 2018-01-12 21:22:40 What linux distro first started to use apk? 2018-01-12 21:33:49 is Alpine only for android is can it be used on PC? 2018-01-12 21:36:37 Alpine's APK has nothing to do with Android's APK. 2018-01-12 21:38:14 KingParrot: you seem to be in the wrong channel. you can ask questions on how to use Alpine in #alpine-linux on Freenode 2018-01-12 21:39:51 I am currently in that chat room and no one is chatting 2018-01-12 21:40:05 many if these irc chat rooms are pretty dead. 2018-01-12 21:41:07 and? ask your questions there, please. this is not really the appropriate forum for most of them. 2018-01-12 21:41:22 I am chating about Alpine in a lubuntu chat room sorta funny 2018-01-12 21:49:36 I think I might try to track down a older version of Adelie to attempt to install. 2018-01-12 21:51:24 KingParrot, this channel is about this linux distro: https://www.alpinelinux.org/ 2018-01-12 21:52:14 alpinelinux has used its own binary package format with the .apk extension since 2006 2018-01-12 21:52:41 as stated, it has nothing to do with the Android package format (unfortunately, also with an .apk extension) 2018-01-12 21:54:55 What is this Linux Vinila? 2018-01-12 21:55:57 Looks like 2011-05-06 Alpine 2.1.5 released is the first one. 2018-01-12 21:56:54 oh intresting so vanilla is a kernel? 2018-01-12 21:56:55 Alpine Linux came standard with GRSEC hardening patches. "Vanilla" version is without those patches, so the kernel is closer to what you would get with Ubuntu or Debian 2018-01-12 21:57:54 Are u a developer on the project? 2018-01-12 21:58:49 I maintain a few packages, and have been involved with the project for some time. But I don't consider myself a developer. 2018-01-12 21:58:51 Mini root filesystem 2018-01-12 21:59:20 What other distros have u worked on? 2018-01-12 21:59:38 You are in a work room. 2018-01-12 21:59:43 Mini: just enough to get the system installed and get a wired connection to the internet. No "extra" packages, like web proxies, database servers, samba, etc. 2018-01-12 21:59:45 Where people are working in silence, think a library. 2018-01-12 22:00:02 What would you say if someone barged into the library and started asking questions better answered outside? 2018-01-12 22:00:17 Please ask your questions in the proper channel. 2018-01-12 22:00:44 what is a devel? 2018-01-12 22:03:48 Origin: Norway 2018-01-13 03:50:25 Is there anyway to override the abuild behavior of where the .apk ends up at the end of the build process? I would like it to stay in the same directory as APBUILD. The only reference I can find for it's location is /etc/abuild.conf but there are still two directories that get appended "[repository]/[architecture]" to the end of $REPODEST, ie. '/custom/x86_64/' 2018-01-13 03:51:31 errr same directory as APKBUILD.. 2018-01-13 20:19:29 https://github.com/alpinelinux/aports/pull/3060 2018-01-13 21:28:04 kaniini: i think we need a symlink /boot/vmlinuz -> /boot/vmlinuz-vanilla 2018-01-13 23:45:50 Natanael Copa: yes, seems reasonable 2018-01-14 17:09:51 jirutka: in case you're still thinking about this: for unrelated reasons I'm soon going to release a version of mdevd that won't need pipeline (it will be a single "mdevd" call, you won't need "mdevd-netlink | mdevd"), so don't worry about it 2018-01-14 17:18:03 skarnet: that’s great! :) 2018-01-14 17:18:19 skarnet: I still didn’t have time to finish that 2018-01-14 19:37:08 Is the current stable kernel in alpine patched against meltdown? 2018-01-14 19:38:19 bulldo00zer: not yet (in v3.7), but we're working on it, it should be soon 2018-01-14 19:38:46 jirutka: txs :) 2018-01-15 00:26:24 long time no see 2018-01-15 13:36:42 hi 2018-01-15 13:54:21 <_ikke_> Hey 2018-01-15 15:48:11 rnalrd: do you have experience with iscsi (SCST and LIO)? or was it fcolista? 2018-01-15 15:48:22 it was me 2018-01-15 15:48:43 although not using it anymore 2018-01-15 15:55:55 ncopa: I configured iSCSI clients on Gentoo few years ago 2018-01-15 15:57:36 bbl, i ask more about iscsi later 2018-01-15 16:05:04 Lio here 2018-01-15 18:58:58 is it possible to convert an old SCST setup to LIO? 2018-01-15 18:59:59 eg, old alpine running SCST. upgrade it and migrate to LIO without reconfigure the clients (or initiators?) 2018-01-15 19:51:45 Hi 2018-01-15 19:53:56 i have a question for you 2018-01-15 19:55:52 :) 2018-01-15 20:00:34 hello 2018-01-15 20:00:42 that's not a question. 2018-01-15 20:02:02 <_ikke_> TimeRot: You don't have to wait for someone to reply 2018-01-15 22:28:39 Hi there! I was just poking around some of the apk source (database.c and package.c) to determine how package checksums (which appear in the APKINDEX file) are generated. I noticed that .apk archives seem use 'pax' extended attributes to store checksums of individual files, but haven't quite tracked down where package checksums come from. Any hints? 2018-01-16 00:37:46 ice799: abuild-tar 2018-01-16 15:24:30 kaniini: I don't think abuild-tar is the right place to look. I took a peak there, looks like the tar checksum that is computed is only 8 bytes wide. The checksum which appears in APKINDEX is a base64 encoded SHA1. 2018-01-16 15:52:22 kaniini: ya, looks like this code from abuild-tar https://github.com/alpinelinux/abuild/blob/master/abuild-tar.c#L285-L363 is responsible for generated the pax headers I found yesterday that contain the checksums of the individual files within an APK 2018-01-16 15:53:27 I am looking for the APK checksum - the one which is base64 encoded and present in the APKINDEX file. I think this checksum is a sha1 of the gunzip'd APK data, but unclear. 2018-01-16 17:48:52 ice799: apk index handles that one :) 2018-01-16 17:56:01 ice799, the sha1 is sha1 over the "control.tar.gz" part of the .apk package 2018-01-16 19:30:14 fabled: OK, thanks. I'll try to figure out where that lives in the APK. So far, I've just seen gzip'd tar files that contain the files to be written, a .PKGINFO, sometimes a digital signature (e.g. .SIGN.RSA*), and PAX headers with per-file sha1s. 2018-01-17 13:54:14 kaniini: fabled: haven't tested this yet, but does it look okay? https://txt.shiz.me/NDA3ZTBkN2 2018-01-17 14:11:41 Shiz: is there any particular reason you do in dev() function even though you're changing content of main package? 2018-01-17 14:11:59 przemoc: because else it gets copied into -dev 2018-01-17 14:12:01 that's the main issue 2018-01-17 14:12:10 it copies it back from dev into the main package 2018-01-17 14:12:12 err, moves* 2018-01-17 14:13:48 I understand what it does, but conceptually it somehow feels wrong to do it that way. 2018-01-17 14:14:01 that's why i checked :P 2018-01-17 14:14:20 i'm not sure how to do it otherwise as i can't presume how $subpkgdir is determined from the main package() 2018-01-17 14:29:37 quickly looking into abuild confirms that your solution may be indeed the easiest one. default_dev has hardcoded moving of various stuff, so to avoid moving headers you'd need to copy-paste whole default_dev and remove include-related parts, which would be much less readable 2018-01-17 14:34:26 of course you could have clang-tailored own dev without meta-ness of default_dev, but then you'd be throwing away existing building blocks and would ask for possible maintanence problems in future coming from diverging from abuild default implementation here 2018-01-17 14:38:24 maybe abuild could support opting-out from some default moves regarding some subpackages, but it's possibly rarely needed, so not sure if it's worth it 2018-01-17 15:45:17 Shiz: yes, it seems fine to me 2018-01-17 16:51:49 przemoc: this is actually kind of the preferred way to override some behaviours of the default functions. 2018-01-17 20:42:38 hi it would seem https://github.com/alpinelinux/aports/commit/1f6e64f5d412c09c7e07b45b3d263c2d7eafc594#diff-d21f431da8c1324fae72aa90228787e6 that is in 3.7 is not in master... 2018-01-17 20:50:37 there's a PR pending for 4.9.1 -> 4.10, that's probably why. 2018-01-17 20:53:19 I am working on XPTI for 4.9.1 2018-01-17 20:54:25 hope that can enter before as its a bit urgent 2018-01-17 21:05:06 https://github.com/alpinelinux/aports/pull/3071 2018-01-17 23:17:11 i managed to bootstrap alpine for mips64/malta qemu. my observations: 2018-01-17 23:17:11 1. lots of packages needed 'file' to be available in order to build shared libs due to: https://lists.gnu.org/archive/html/libtool-patches/2013-08/msg00001.html 2018-01-17 23:17:14 2. ncurses tarball was removed upstream, but there is a copy available at: http://bld1.alpinelinux.org/distfiles/v3.6/ncurses-6.0-20170930.tgz 2018-01-17 23:17:17 3. latest linux-vanilla needed a bit of tweaking to support cross-compiling again 2018-01-17 23:17:20 (i skipped building go and ghc) 2018-01-18 14:12:04 wishlist: any chance of getting pdmenu packaged? :) 2018-01-18 14:19:13 <_ikke_> If you ask nice enough :P 2018-01-18 14:20:27 :D i wouldnt want it to be a burden; if noone else has asked for it ... 2018-01-18 14:20:55 i must first finally get the building vm up; so i'll be able to cross-compile things myself 2018-01-18 19:08:31 clandmeter2: I just remembered that you pinged me week ago, but I wasn't around back then and forgot to ask later how could I help you 2018-01-18 19:18:22 dsabogal: ghc cross compiling is definitely optional and non fun i don't blame you :) 2018-01-18 19:25:17 ghc is only built for x86_64 and armhf anyway 2018-01-18 19:52:59 atm 2018-01-18 19:53:26 i'm working on the aarch64 and x86 ports, slowly, had to deal with comcast and 20k/s speed this past week 2018-01-18 19:57:59 mitchty: oh god, I feel ya. in december both my DSL lines decided to sync at 200 Kbit/s or not at all 2018-01-18 19:58:04 it was not a pleasant experience 2018-01-18 20:01:31 i had git clone timeout, using the internet now at 20k/s is... almost impossible 2018-01-18 20:03:12 this is how i learnt there is a 10 minute timeout in git i was unaware of :) 2018-01-18 20:07:40 heh, pretty sure I also encountered that timeout for the first time in that period 2018-01-18 20:08:41 luckily there's decent lte coverage here, so I ended up tethering with my phone most of the month 2018-01-18 20:08:44 that phone bill tho 2018-01-18 20:35:34 dsabogal: sounds like you have done an awesome work with mips64! 2018-01-18 20:42:38 ncopa: thanks. it was mostly a follow-up to https://bugs.alpinelinux.org/issues/5643 2018-01-18 20:43:04 i have an octeon machine that can serve as builder 2018-01-18 20:43:07 16gb ram 2018-01-18 20:43:29 i just need to know what to do to make it run 2018-01-18 20:43:56 sounds like dsabogal has the core packages needed to get new builder up 2018-01-18 20:44:21 dsabogal: do you think it would make sense to add 'file' as a build-base dependency? 2018-01-18 20:44:42 i think so 2018-01-18 20:44:52 we have been snagged by that with adelie 2018-01-18 20:46:36 dsabogal: could you please file an issue on bugs.a.o requesting adding 'file' as a build-base dependency? So I can refer to it in the commit message 2018-01-18 21:01:26 kaniini: ncopa: any chance you can have a look at https://github.com/alpinelinux/aports/pull/3071 today? 2018-01-18 21:01:43 I am working on alpine-3.5 (XEN 4.7.3) now as wel 2018-01-18 21:02:55 HRio: i am working on 4.14.14 kernel as we speak. I'll look at that directly after 2018-01-18 21:03:06 thanks 2018-01-18 21:03:59 HRio: is that PR for 3.7-stable or git master? 2018-01-18 21:04:54 master 2018-01-18 21:05:03 not done 3.7 yet 2018-01-18 21:05:16 ok. lets do master first 2018-01-18 21:05:38 i need to go out a bit later but i think I'll reach to have it merged today 2018-01-18 21:05:42 i'll do my best 2018-01-18 21:05:43 I do not have any Intel hypervisors on 3.7 2018-01-18 21:06:09 so working on 3.5 and after that 3.6, to get some Intel testing 2018-01-18 21:08:51 ok. thank you for working on it 2018-01-18 21:08:59 appreciate your help 2018-01-18 21:17:01 glad to be of some help, you guys do so much for me, everyday. 2018-01-18 21:24:11 kaniini: i think the steps you need are: 2018-01-18 21:24:11 1. these patches i sent to the ml for linux-vanilla (small change for mips64, and fixing cross-compile) 2018-01-18 21:24:14 2. add a kernel config for octeon board (i used buildroot's config for the malta) 2018-01-18 21:24:17 3. apk add file (i personally just added it to makedepends but a lot of packages ended up needing it) 2018-01-18 21:24:20 4. edit /usr/share/abuild/functions.sh https://git.alpinelinux.org/cgit/abuild/commit/?id=55cef04cf9daf5d9765860cd4bb4be1e0118daf5 2018-01-18 21:24:23 5. cd aports; ./scripts/bootstrap.sh mips64 2018-01-18 21:24:25 6. if linux-vanilla install fails due to missing vmlinuz fails, you might need this: https://www.linux-mips.org/archives/linux-mips/2018-01/msg00352.html 2018-01-18 21:26:40 (for step 5, you can skip building 'go' and 'ghc' by commenting those packages out. and perhaps add 'file') 2018-01-18 21:37:37 looks like update-kernel script is broken 2018-01-18 21:37:38 or it is the mkimage.sh script thats broken 2018-01-18 21:44:26 ncopa: 3.5 XEN-XPTI https://github.com/alpinelinux/aports/pull/3074 2018-01-18 21:45:39 ncopa: update-kernel sometimes does "fun things" if /tmp runs full while its doing its stuff, but probably unrelated to you problem 2018-01-18 21:51:01 i think its due to -vanilla suffix rename 2018-01-18 21:51:02 im investigating 2018-01-18 21:51:12 need fix it so i can test boot in qemu before i push 2018-01-18 21:51:27 ah it probably expects to find /boot/vmlinuz 2018-01-18 21:51:59 $ ls /tmp/update-kernel.N9v9M8/root/boot/ 2018-01-18 21:51:59 System.map-vanilla config-vanilla vmlinuz-vanilla 2018-01-18 21:52:08 i think its the System.map that should not be renamed 2018-01-18 21:52:55 it was System.map-hardened before 2018-01-18 21:53:08 so it should be System.map-vanilla now, no? 2018-01-18 21:53:17 ok, the its the update-kernel script that broke 2018-01-18 21:53:34 i need to go out for a couple of hours 2018-01-18 21:53:35 bbl. 2018-01-18 23:21:54 kaniini: the config used here could possibly be useful for creating a linux-octeon package, which we sorely need 2018-01-18 23:21:54 https://github.com/xypron/kernel-edgerouter 2018-01-18 23:41:43 i prefer to do it as linux-vanilla subprofile instead of separate APKBUILD but yes :) 2018-01-19 08:23:14 kaniini: ahh, I were not aware you could do that :) 2018-01-19 08:27:09 so I'd need to add config-octeon.mips64 right? 2018-01-19 12:23:29 any github experts around? is it possible to use the github issue search to find all issues that do not have a label? 2018-01-19 12:26:40 ok, found it. for the record: `is:open no:label` 2018-01-19 14:06:45 Lochnair: yes, absolutely 2018-01-19 19:16:12 ncopa: I am done with XPTI for alpine-3.6 now as well: https://github.com/alpinelinux/aports/pull/3078 2018-01-19 19:17:13 HRio: do we have XPTI for alpine-3.7 too? 2018-01-19 19:18:32 this one is for 3.7-stable right? https://github.com/alpinelinux/aports/pull/3077 2018-01-19 19:19:36 very good 2018-01-19 19:19:48 i will try merge those later tonight 2018-01-19 19:19:51 no its was in 3.7 already 2018-01-19 19:19:57 but missed in edge 2018-01-19 19:20:02 aha 2018-01-19 19:20:58 let me merge that then before i forget it 2018-01-19 19:21:35 I will try to spend some time on XPTI for 3.7 this weekend as well, lets see... 2018-01-19 19:22:13 is XEN in 3.4 still within our security support? 2018-01-19 19:22:25 yes 2018-01-19 19:23:37 xen 4.6 right? 2018-01-19 19:24:37 there are XPTI backports available for XEN 4.6 -> 4.10 2018-01-19 19:24:53 i think we should have the CVE is in the secfixes comment 2018-01-19 19:24:53 ok i need to go 2018-01-19 19:24:54 i'll try finish it up later tonight 2018-01-19 19:26:11 ok, thanks. 2018-01-19 19:27:43 ncopa: (for later) the CVEs are here http://xenbits.xen.org/xsa/advisory-254.html 2018-01-19 19:31:40 would you like to take over xen maintainership? :p 2018-01-19 19:59:46 kaniin: ;-) no, but I will try to help out more often in the future 2018-01-19 20:02:15 noticed a bug in /etc/xen/scripts/block-iscsi is expects lun0 always, but with tgtd lun-0 is the controller and lun-1 is the disk (http://inqbus-hosting.de/support/dokumentation/docs/target-daemon-tgtd-tgtadm#show-targets) 2018-01-19 20:04:57 fixing that iscsi works perfectly directly in the domU cfg ( 'script=block-iscsi,vdev=xvda1,target=iqn=iqn.2015-08.se..:-disk,portal=,xvda1,w' ) 2018-01-19 20:15:02 i am working on creating a virtualization interest group that will take over maintenance anyway :) 2018-01-19 20:17:25 yes, its a lot of testing when doing changes like XPTI... 3.4, 3.5, 3.6, 3.7 and edge 2018-01-19 20:18:46 I still have one XEN dom0 on 3.5, but none on 3.4. 2018-01-19 20:32:01 (i also intend to charter a Security working group, to shut up the people who say alpine sucks because we do not have one) 2018-01-19 20:34:30 :-) 2018-01-19 20:35:18 I switched dom0's from an other dist to Alpine, cause security fixes for Alpine XEN was much faster... 2018-01-19 20:44:29 kaniini: I testbuilt https://github.com/alpinelinux/aports/pull/3026 and it builds fine in my edge devel domU. It seems from your comment that it failed for you? 2018-01-19 20:45:25 having said that, I have bash installed... 2018-01-19 20:46:02 but the package seems to depend on bash anyhow... 2018-01-19 22:08:32 kaniini: I made an attempt at adding an octeon flavor for linux-vanilla, but it fails at the packaging stage because there's no vanilla flavor for mips64 2018-01-19 22:08:38 current patch here: http://dpaste.com/34ZB3ZR 2018-01-19 22:11:19 it's based on the virt-x86_64 config, so there's likely a lot of unnecessary drivers in there 2018-01-19 22:11:31 but I wanted to get it to build first before combing over the config 2018-01-19 22:44:04 kaniini, is there any ongoing discussion regarding that? 2018-01-19 23:02:56 https://pkgs.alpinelinux.org/ shows a 502 bad gateway 2018-01-19 23:03:11 <_ikke_> clandmeter: might be working on it 2018-01-19 23:03:58 :) 2018-01-19 23:04:07 can never do anything without ppl noticing 2018-01-19 23:04:11 <_ikke_> hehe 2018-01-19 23:05:09 im migrating it so please wait a min 2018-01-19 23:05:45 clandmeter: okay, thanks :) 2018-01-19 23:14:14 danieli: ongoing discussion regarding which 2018-01-19 23:14:25 Lochnair: hmm lets just use fake vanilla config for the moment 2018-01-19 23:14:49 kaniini: regarding a security working group 2018-01-19 23:15:46 Supposedly there's a room on the Matrix server where build servers report successes and failures? I thought it was #commits but that and #commit both say room not found. 2018-01-19 23:15:56 AWilcox[m]: #alpine-commit 2018-01-19 23:16:03 Is there a list of Matrix channels on the Alpine Linux server? Or can someone point me to the correct room name? 2018-01-19 23:16:05 dammit, I mean #alpine-commits 2018-01-19 23:16:17 Ahh, okay 2018-01-19 23:16:29 it's a freenode channel, but should be accessible through matrix 2018-01-19 23:16:56 Thanks! It works. 2018-01-19 23:17:22 the build system is hooked up to algitbot through MQTT, which is set to post in #alpine-commits 2018-01-19 23:29:59 danieli: it is in progress. we want generic framework for chartering WGs/SIGs first 2018-01-19 23:30:36 alpine infra has strange status 2018-01-19 23:30:46 it indeed seems to 2018-01-19 23:31:11 so the idea is that alpine infra will just be a WG like any other 2018-01-19 23:31:37 it isn't really more than a WG though, but it's kind of... obscure, at this point 2018-01-19 23:31:52 part of this is already implemented: sysadmins now have guaranteed equal rights in the project 2018-01-19 23:31:58 because everyone is now a subclass of Member ;) 2018-01-19 23:32:14 yes, i just mean organizationally speaking it is weird 2018-01-19 23:32:33 in practice, it is already "just a WG" 2018-01-19 23:32:41 what defines a sysadmin in this case? 2018-01-19 23:33:00 but we want to sort the remaining governance issues around WGs before creating anymore 2018-01-19 23:33:07 it is being worked on, however 2018-01-19 23:33:17 also, that email you just sent to alpine-devel describing member status - won't giving every member an email account be quite the hassle? 2018-01-19 23:33:29 or does alpine now have someone dedicated to being postmaster and hostmaster? 2018-01-19 23:33:56 the alpine-devel mail covers that. in essence, it is whoever has been granted admin rights to our systems 2018-01-19 23:34:05 that's the infra team's job :) 2018-01-19 23:34:11 kaniini: thanks for writing down the e-mail, that must have been a lot of work. will this be added to the wiki somewhere? I think it would be easier to find there 2018-01-19 23:34:19 gotcha, everyone on the infra team is a member with some kind of access to some system 2018-01-19 23:34:23 ollieparanoid: we are going to put it on the website 2018-01-19 23:34:43 yes, membership is implied by having root ;) 2018-01-19 23:35:22 after all we would not want to have somebody who is not a member of the project with root access to the project's machines 2018-01-19 23:35:31 indeed 2018-01-19 23:35:32 know what i mean? 2018-01-19 23:35:54 that's the infra team's job :) // regarding my hostmaster/postmaster question? if so, who are they, specifically? 2018-01-19 23:35:59 so previously they had implicit kinda/sorta developer status 2018-01-19 23:36:12 I had no idea 2018-01-19 23:36:26 now they are just members who do not have commit 2018-01-19 23:36:27 unless they do the maintainer 2018-01-19 23:36:28 track 2018-01-19 23:36:29 stupid mac keyboard 2018-01-19 23:36:31 I always saw the "infra team" as a diffuse kind of working group 2018-01-19 23:36:34 dunno, ask the infra team ;) 2018-01-19 23:36:47 (probably clandmeter if i were to guess) 2018-01-19 23:36:52 according to wiki.intra.a.o, I am on the infra team 2018-01-19 23:37:01 yes,like i said, weird status 2018-01-19 23:37:10 it's just not synchronized at all 2018-01-19 23:37:15 we are trying to more concretely define this 2018-01-19 23:37:25 defining clear roles doesn't necessarily define clear duties regarding specific systems and services 2018-01-19 23:37:27 either way, infra team reports to core team 2018-01-19 23:37:55 indeed, but from my end we just delegate that to "you guys" 2018-01-19 23:38:38 i have made a mental note that we should probably audit all of that mess though 2018-01-19 23:38:45 so we can figure out who is doing what 2018-01-19 23:38:46 and document it 2018-01-19 23:38:46 we really do have to audit it 2018-01-19 23:38:48 I'm thinking of making some changes to wiki.intra - making a table of *all* developers, sysadmins and contributors and clarifying their duties 2018-01-19 23:39:26 I suppose clandmeter is taking the lead on the infra team 2018-01-19 23:39:33 well, right now, the core team is presently authoritative for who has an account and what kind of account it is 2018-01-19 23:39:55 (we are planning to delegate this as well, to a working group similar to Debian's account managers) 2018-01-19 23:40:05 figured as much 2018-01-19 23:44:06 however, i think that there are some accounts in different statuses 2018-01-19 23:44:12 so we should probably audit this as well 2018-01-19 23:44:16 elaborate? 2018-01-19 23:45:26 folks like Aerdan have old email accounts that predate the non-technical member status 2018-01-19 23:45:34 I see 2018-01-19 23:46:12 ollieparanoid: does it work now? 2018-01-19 23:50:02 yes, pkgs.a.o works now. and I see that it was slightly redesigned. 2018-01-19 23:50:47 yes 2018-01-19 23:50:55 it should match www 2018-01-19 23:51:02 it has less js 2018-01-19 23:51:09 it should be a lot faster 2018-01-19 23:51:37 and the flagged listing should clean it self better. 2018-01-19 23:52:13 meaning if one builder has updated the pkg it should be removed from flagged page. 2018-01-19 23:52:53 yeah.. some builders don't stay up to speed with the others 2018-01-19 23:53:01 that always annoyed me 2018-01-19 23:53:35 we need better buildserver software >.< 2018-01-19 23:54:48 what is it like as of now? 2018-01-19 23:54:51 I have no insight into that 2018-01-19 23:56:14 we have our own set of scripts 2018-01-19 23:56:54 that much I know, isn't everything there a bunch of scripts and tools? 2018-01-19 23:58:42 IIRC fabled was thinking out loudly about switching to buildbot in the past, but there were some shortcommings needed to be addressed before that would happen. 2018-01-19 23:59:21 but I may be remembering incorrectly ;) 2018-01-19 23:59:38 there's a WIP thing called abuildd 2018-01-20 00:00:00 not sure what the progress is on that, though. once that happens I can start working on apkfission.net 's stack. 2018-01-20 00:01:10 nobody will help with abuildd 2018-01-20 00:01:12 and i have limited time in one day 2018-01-20 00:01:25 so in other words it's dead now. great. 2018-01-20 00:01:30 is there a trello board or similar? 2018-01-20 00:01:35 some todo tracker 2018-01-20 00:01:39 not dead 2018-01-20 00:01:44 just need patches 2018-01-20 00:01:56 bugs.a.o is kind of cluttered 2018-01-20 00:02:24 Aerdan: i might work on it in february after i finish replacing this security hole activitypub shit i have installed 2018-01-20 00:02:44 okay 2018-01-20 00:03:45 Aerdan: patches accepted tho 2018-01-20 00:04:02 i mean, in theory, this stuff is not that hard to write 2018-01-20 00:08:03 kinda hard to write patches without code *to* patch ;) 2018-01-20 00:11:26 danieli: it indeed is. maybe I'll try to repeat my cleaning attempts, as I did a few years back. (re)categorization, better naming, marked as fixed if I believe it is, providing fix myself if it seems a quick matter, etc. also when OP is no longer interested in its own tickets for instance, it's not that wise for us to keep them forever. 2018-01-20 00:11:58 in the past there was discussion about improving categories in bugs.a.o, but it ended without any conclusion, or at least w/o any action 2018-01-20 00:12:06 przemoc: I can help out a bit with being a janitor there, perhaps designing an alpine theme for redmine 2018-01-20 00:12:24 I can shoot an email to alpine-infra to spur conversation about cleaning up bugs.a.o 2018-01-20 00:12:29 aerdan: you know python :) 2018-01-20 00:12:34 leaving it for dead here in IRC won't help 2018-01-20 00:12:55 kaniini: what exactly is needed? 2018-01-20 00:14:17 https://github.com/kaniini/abuildd 2018-01-20 00:14:25 collect/compose/enqueue/monitord 2018-01-20 00:14:34 other than enqueue they are obvious 2018-01-20 00:14:47 (!) Issues (0) 2018-01-20 00:14:51 enqueue is tricky because it needs to parse APKBUILDs to build the dependency graph 2018-01-20 00:15:06 I see, so you need it written, and not just fixed? 2018-01-20 00:15:14 anyway, patches for any of those accepted :) 2018-01-20 00:15:17 yes 2018-01-20 00:15:28 i had to set it aside and work on governance 2018-01-20 00:15:31 and policy docs 2018-01-20 00:15:37 gotcha 2018-01-20 00:15:44 oh, you're using asyncio, neat 2018-01-20 00:15:50 after reading kaniini's mail I also think I should somehow officially ask to become AL member, because I always felt member-ish, even while not being a member, but OTOH I possibly haven't contributed that much yet to backup my member-ish status ;) 2018-01-20 00:15:56 so that we can have more orderly functionality of the project 2018-01-20 00:15:57 what python version (minor) is this targeting? 2018-01-20 00:16:13 przemoc: send me some examples 2018-01-20 00:16:13 danieli: 3.6 2018-01-20 00:16:42 przemoc: but overall, yes i would agree at least for non-technical (if you do not wish to do the technical track) 2018-01-20 00:16:46 my favorite 2018-01-20 00:16:57 danieli: i also intend to use uvloop 2018-01-20 00:17:03 why? 2018-01-20 00:17:05 GOTTA GO FAST afterall 2018-01-20 00:17:14 replacement for the default crap abstracteventloop? 2018-01-20 00:17:18 yes 2018-01-20 00:17:22 you dont know it? 2018-01-20 00:17:27 it is asyncio eventloop on libuv 2018-01-20 00:17:37 i figured as much considering the "uv" in "uvloop" 2018-01-20 00:17:41 i remember old pyuv 2018-01-20 00:17:44 it is literally 2018-01-20 00:17:53 10x faster than asyncio stock 2018-01-20 00:17:56 and supports 0-copy io 2018-01-20 00:18:01 fancy stuff 2018-01-20 00:18:08 neat, i'll familiarize myself with uvloop then, thanks 2018-01-20 00:18:11 shouldn't be hard, looks trivial 2018-01-20 00:18:27 yep, it is trivial, at least the abstraction you use 2018-01-20 00:19:30 like i said, the only hard thing is the depsolver 2018-01-20 00:19:30 i just have this thing 2018-01-20 00:19:40 where things that are shit simple 2018-01-20 00:19:49 get set aside 2018-01-20 00:19:54 for things that are more important 2018-01-20 00:20:07 yep, that's how priorities work 2018-01-20 00:20:10 just needs some manpower i guess 2018-01-20 00:21:50 it reminds me i need to fix the provides issue i have now in pkgs.a.o 2018-01-20 00:23:57 now all lua packages seems to depends on all lua versions :| 2018-01-20 01:11:52 So the armhf builder is failing to build lz4; SIGBUS while running `frametest`. I'm not seeing any issues (it passes and builds successfully) on my ARM device. How can I help debug it? 2018-01-20 01:12:19 I'm using a Raspberry Pi (original model 1 B) locally for this test 2018-01-20 01:12:45 And a brand new chroot of Alpine Edge 2018-01-20 01:46:19 AWilcox[m]: SIGBUS kinda sounds like a memory alignment issue 2018-01-20 01:48:19 I know it does, but it isn't happening here with the same APKBUILD :/ 2018-01-20 11:36:04 we have a bunch of security related xen PRs on GitHub https://github.com/alpinelinux/aports/pulls?q=is:pr+is:open+label:T-security could someone actually using xen take a look at these? 2018-01-20 11:36:48 that's something that needs to be prioritized 2018-01-20 11:38:37 nmeum: core team's setting up a security team in part to make people quit complaining that we don't have one. I think when that's done those PRs will be handled. 2018-01-20 11:39:17 there's plenty of red tape before a security team can be assembled 2018-01-20 11:39:44 and security is important, red tape or not 2018-01-20 11:40:23 I suppose I'm not in a position where I have a right to speak up, but it is important 2018-01-20 11:41:13 it is, yes, but so is having a group of people specifically keeping track of security concerns. 2018-01-20 11:43:57 my point is that *someone* should handle security issues (as usual) until a security working group is established 2018-01-20 11:55:21 heh, implementing processes is fun 2018-01-20 11:56:27 especially if it takes us from the distro with the fastest sec fixes to being super slow. hopefully this will be back to normalsoon.... 2018-01-20 12:26:59 could someone give me some feedback on https://github.com/alpinelinux/aports/pull/3076 jirutka could you take a quick look at this? 2018-01-20 13:01:17 kaniini: not quite sure what you mean by "fake" kernel config 2018-01-20 13:51:01 Lochnair: i think kaniini means having config-vanilla.mips64 be the kernel config for the octeon (for now) 2018-01-20 14:55:27 dsabogal: won't that get confusing later? 2018-01-20 18:17:47 it can be moved later. 2018-01-20 18:57:38 nmeum: I tried to fix the lz4 thing but it doesn't crash on my armhf builder running edge. Nobody is giving me a backtrace or core or anything so I can try to fix it. 2018-01-20 19:13:18 hm, have you asked clandmeter2? 2018-01-20 19:13:27 or ncopa? 2018-01-20 19:13:46 there is also a GitHub PR upgrading lz4, maybe it fixes this issue? 2018-01-20 19:59:07 AWilcox[m]: please use license names from SPDX License List 3.0 when you're fixing them. GPL-2.0+ is old and depracted name (from previous version), GPL-2.0-or-later is the correct one now. 2018-01-20 20:01:01 Adélie's standardised on 2.6, because 3.0 is overly verbose. (I asked roughly when SPDX 3.0 came out.) 2018-01-20 20:04:30 I agree it's somewhat verbose, but it's also more clear and matches what is used in actual licenses, so it's kind of more human-friendly too. 2018-01-20 20:05:14 SPDX 3 came out at the end of last year, true, but before next AL will be relased, it will be almost half a year already, so I don't think it's unreasonable to target it instead of older version. 2018-01-20 20:05:26 s/rela/relea/ 2018-01-20 20:58:16 How to promote security fixes for old releases? Ci fail on it https://github.com/alpinelinux/aports/pull/3088 2018-01-20 21:00:52 dsabogal: did the sha512sum in the ncurses APKBUILD match with the archive on the mirror you found? 2018-01-20 21:21:55 in my opinion, we should replace ncurses with BSD ncurses 2018-01-20 21:22:12 it is a much simpler implementation, so is probably better from security POV 2018-01-20 21:24:05 agreed 2018-01-20 21:53:11 Guys, is fetching upstream patches during build() or prepare() in APKBUILD allowed? 2018-01-20 21:54:01 I don't want mess git repo via some big patches 2018-01-20 21:55:59 Include them as files. 2018-01-20 21:56:26 but it is going to be stored in /var/cache/distfiles 2018-01-20 22:21:20 Lochnair: it does 2018-01-20 23:06:29 That's going to result in a tonne of merge conflicts. 2018-01-20 23:06:56 I'm not sure how much we'll be able to share with Alpine if SPDX 3 is being used :/ 2018-01-20 23:32:47 hm, that sucks 2018-01-21 09:09:04 dyson 2018-01-21 09:09:28 Opps, sorry. Don't mind me. 2018-01-21 22:22:08 MATE users: Did anyone experience strange artifacts in some applications with a recent update in Alpine edge? Here are two photos of the same device running XFCE4 and MATE, and the application matchbox-keyboard is not rendering correctly in MATE: https://github.com/postmarketOS/pmbootstrap/issues/1149 2018-01-21 23:01:23 MATE went to gtk3 during the alpine 3.7 cycle 2018-01-21 23:15:43 thanks, that sounds very related 2018-01-22 15:51:47 got an error with : 2018-01-22 15:51:47 Error relocating /usr/lib/python3.6/site-packages/PyQt5/QtCore.so: _ZN23QOperatingSystemVersion11AndroidOreoE: symbol not found 2018-01-22 15:51:52 https://dpaste.de/zUoi 2018-01-22 15:52:17 py3-qt5 2018-01-22 17:00:59 py3-qt5 probably need rebuild 2018-01-23 05:18:06 I am trying to patch an apk. How do I install it and test it? I did an abuild -r, but I can't test it. 2018-01-23 05:34:14 <_ikke_> koldbrutality: it adds them to a repository located in ~/packages 2018-01-23 05:34:46 thx 2018-01-23 06:07:16 Guys, can I assign a separate $pkgver for a sub-package ? 2018-01-23 06:09:03 <_ikke_> Would that even make sense? 2018-01-23 06:10:17 Maybe a plugin that has a separate distribution file? 2018-01-23 06:10:31 thus, separate version.. 2018-01-23 06:12:24 <_ikke_> I don't think it's expected 2018-01-23 06:14:43 I have to increase pkg release number to trigger new build but whole package is going to be rebuild. 2018-01-23 06:15:26 so, separate version only viable if subpackage is is also separate.. 2018-01-23 06:16:10 I'm talking about testing/vdr 2018-01-23 06:17:35 btw, should I file a request to move a package from testing to community? 2018-01-23 06:17:56 who is deciding which package is fine to move other repos? 2018-01-23 06:19:56 <_ikke_> The maintainer can request it 2018-01-23 06:20:15 <_ikke_> One of the core devs will decide if it makes sense 2018-01-23 06:21:06 _ikke_: thanks mate 2018-01-23 07:24:04 <_ikke_> One of the core devs will decide if it makes sense 2018-01-23 07:24:06 <_ikke_> oops 2018-01-23 07:32:30 gg 2018-01-23 07:47:35 What does this mean? https://travis-ci.org/alpinelinux/aports/builds/332175163?utm_source=github_status&utm_medium=notification 2018-01-23 07:47:58 It works fine on my machine 2018-01-23 07:48:39 And it takes several minutes to build it more than 6 2018-01-23 07:48:44 min 2018-01-23 07:49:32 because it has to compress/optimize those emojis 2018-01-23 07:49:55 I will time it 2018-01-23 09:06:15 koldbrutality: bundle those depended commits in one pr 2018-01-23 09:06:41 you cannot depend on something that is not available in current repo 2018-01-23 09:10:10 I should close all those pull requests? 2018-01-23 09:10:37 yes, close them and move the commits to the emoji branch 2018-01-23 18:04:49 ping on https://github.com/alpinelinux/mkinitfs/pull/21 2018-01-23 18:53:00 mmlb: thanks. Iooks good. I'll merge it 2018-01-23 18:59:45 +1 2018-01-23 19:15:29 hi 2018-01-23 19:40:50 hi 2018-01-23 19:43:34 <_ikke_> hi 2018-01-23 19:46:17 <_ikke_> Would be nice if someone could take a look at https://github.com/alpinelinux/aports/pull/2433 2018-01-23 20:25:06 You can't have alternative choices for APKBUILDs right? Like choice between SSL library or choice between FFMpeg or libva. I know it is possible on Gentoo's ebuilds. 2018-01-23 20:25:58 such things are not possible 2018-01-23 20:26:13 if you want capabilities like that, you should use gentoo ;) 2018-01-23 20:26:32 I quit Gentoo because I don't have time. 2018-01-23 20:43:00 well, the only way to have your choices is by doing it yourself. Gentoo automates that. We, as a binary-oriented distro, don't and generally can't. 2018-01-23 20:50:07 yeah, not easily doable at the level Gentoo has it. I remember there was some talk about a gentooish mechanism for Alpine for choosing some default tools but I can't remember either the Gentoo feature corresponding to that or what was said in that discussion 2018-01-23 20:50:36 I have, thank goodness, started to forget details of Gentoo... 2018-01-23 20:51:34 <_ikke_> 160 PRs open right now on github 2018-01-23 21:51:48 _ikke_: http://build.alpinelinux.org/buildlogs/build-edge-ppc64le/testing/fzf/fzf-0.17.0.2-r0.log should i just blacklist those archs 2018-01-23 22:32:13 You guys think I should submit Firefox's APKBUILD with FFMpeg support, or make a dependency with libva, or not submit it at all. Its a simple fix but could increase security risk. 2018-01-23 22:32:59 It allows to play YouTube live video and some Internet radio. 2018-01-23 22:35:41 The problem is that FFMpeg and libva are like mutual exclusive. Also FFMpeg has more security CVEs in 2017 47 of them compared to Libav has only 24 according to www.cvedetails.com. 2018-01-23 22:35:50 well, is it either ffmpeg OR libva or can it be both? 2018-01-23 22:36:01 wait, are you talking about libva or libav after all? 2018-01-23 22:36:14 oh, you typoed libav. libav is Debian trash, avoid. 2018-01-23 22:37:17 Debian is back on ffmpeg since 9.0 2018-01-23 22:37:21 On Gentoo you can't have both, so I would expect here. Libav according to Firefox bugzilla. 2018-01-23 22:38:12 This is for Firefox-57.0.3 2018-01-23 22:38:51 FFMpeg has more codecs but more security risk if you don't update frequently. 2018-01-23 22:40:07 4 code execution CVEs for FFMpeg compared to 0 for libav in 2017. 2018-01-23 22:47:00 well, that's probably because nobody cares about libav any more. :) 2018-01-23 23:16:21 Should I make a pull request as a seperate package (testing/firefox-h264) or keep it in the same package (testing/firefox)? 2018-01-23 23:20:13 let me see if I can just delete ffmpeg and it will still function 2018-01-23 23:24:53 It still works. I think I just keep it in the same package. If they don't want h264 support, they can just delete ffmpeg-dev and ffmpeg-libs. 2018-01-24 00:03:41 koldbrutality: can be firefox-h264 built separately from Firefox? 2018-01-24 00:03:59 koldbrutality: is it like a bundled dependency of ff, or is it part of the ff? 2018-01-24 00:05:24 It would be an entire copy of testing/firefox but with FFMpeg support. 2018-01-24 00:06:28 So mutually exclusive 2018-01-24 00:06:31 aha 2018-01-24 00:06:50 I don’t think we want to maintain two copies of firefox package… 2018-01-24 00:07:06 or to build firefox twice 2018-01-24 00:07:12 is it really needd? 2018-01-24 00:07:25 I know that's why we just tell them to delete ffmpeg if they don't like the risk. 2018-01-24 00:08:34 sry that i didn’t read full backlog, but how it is – FF cannot play H.264 videos at all w/o ffmpeg, or just don’t support some less used codecs? 2018-01-24 00:10:12 It could be because of licensing or patents. 2018-01-24 00:11:14 OpenH264 Video Codec provided by Cisco Systems, Inc. 2018-01-24 00:11:35 this is plugin I have in FF on macOS 2018-01-24 00:11:47 so it doesn’t look like H.264 is provided by ffmpeg for ff 2018-01-24 00:35:05 What is the difference between depends_dev and makedepends? 2018-01-24 00:35:39 I want to just link to the library then dispose of it after building the package. 2018-01-24 00:55:12 depends_dev is for things needed to develop with the software that aren't build-time dependencies. 2018-01-24 03:20:10 Is there a location where python scripts are installed? 2018-01-24 03:25:22 Probably /usr/share for the appropriate package, unless they're meant to live in /usr/bin. 2018-01-24 03:27:23 I found it. I was trying to find the site-packages (/usr/lib/python{2.7,3.6}/site-packages) directory. 2018-01-24 03:27:33 This package doesn't have a setup.py script. 2018-01-24 03:28:13 which software is this? 2018-01-24 03:28:23 for py-jedihttp 2018-01-24 03:28:46 I was going to make a package for ycmd /YouCompleteMe 2018-01-24 03:29:48 this doesn't seem like it was intended to be packaged, just run 'as is'. 2018-01-24 03:30:23 They keep bundling in ycmd. 2018-01-24 05:29:10 <_ikke_> kaniini: Looks like if go can build for the platform, it's easy enough to patch it it, but for me it would be good enough to blacklist those arches 2018-01-24 05:58:14 _ikke_: i'd be amazed if go didn't build on ppc64le, it's a supported target by upstream 2018-01-24 06:00:25 <_ikke_> Xe: Yeah, I was figuring as mutch 2018-01-24 06:00:56 <_ikke_> The Makefile just lists some of the arches 2018-01-24 08:18:57 <_ikke_> Added this because I could: https://zabbix.alpinelinux.org/history.php?action=showgraph&itemids[]=28584 2018-01-24 08:19:07 <_ikke_> Number of open issues in aports on github 2018-01-24 13:58:32 _ikke_: i need login for that 2018-01-24 14:08:07 <_ikke_> ncopa: You have an account there 2018-01-24 14:08:12 <_ikke_> ncopa: should I reset the password? 2018-01-24 14:08:59 lets take it in #alpine-infra 2018-01-24 14:57:44 since everybody keeps asking, let me ask this a bit differently... is there a wiki entry or a similar readable document on the project's efforts wrt. de-grsecurity and such? 2018-01-24 14:59:06 TBB: at this point, no. 2018-01-24 14:59:06 but yeah, we should write an article about it 2018-01-24 14:59:07 we should probably publish it under news section on alpinelinux.org 2018-01-24 15:01:04 okay; that would be a good idea since, well, at least I would at some point grow tired of the constant questions 2018-01-24 15:02:43 I've had a bit of an interesting day at work because of the Intel mess all of a sudden has again peaked interest in this matter 2018-01-24 15:04:11 I only reported to my superiors in March last year that there are going to be grsecurity related changes coming, and as a reaction to it, fuck all happened. now, fortunately, people woke up 2018-01-24 16:29:53 not too sure how much I want to recommend my customer to get a grsecurity subscription ... professionally it would probably make sense, but the moral questions, ohh those moral questions :) 2018-01-24 16:31:54 Maintainers, could someone please review and perhaps merge https://github.com/alpinelinux/aports/pull/3010. Thanks! 2018-01-24 17:12:17 ncopa: what should we do with the XEN (XPTI) patches? 2018-01-24 17:13:01 Planning to start work on 4.8.3 but would be nice to get the security fixes in before that. 2018-01-24 17:14:41 note 4.8.3 is for alpine 3.6 2018-01-24 18:33:27 How does multiarch work? Will it run build*( twice per arch if I put x86_64 and x86? 2018-01-24 18:35:05 <_ikke_> There is no default multiarch 2018-01-24 18:36:44 So abuild -r runs once for the current arch? 2018-01-24 18:37:48 <_ikke_> yes 2018-01-24 18:38:06 <_ikke_> the arch array just defines what arches the package will support 2018-01-24 18:41:17 If I haven't tested it on an arch I shouldn't add it right? I think this would likely work on x86 but I haven't tried. 2018-01-24 18:42:41 <_ikke_> You can first assume all arches, and restrict it when an arch is not supported 2018-01-24 18:42:56 <_ikke_> note that there are more arches then just x86 and x86_64 2018-01-24 18:43:38 I don't want to give the wrong impression that it only works on x86_64. 2018-01-24 18:44:06 The only thing that it would hold it back is clang and boost 2018-01-24 18:45:22 Hi. I ran into an IO ERROR for mariadb-dev-10.1.28-r1 from nl.alpinelinux.org/alpine/edge/main/x86_64/. What can be done to correct this? 2018-01-24 18:47:29 We need to see the whole log/context. 2018-01-24 18:47:46 sudo apk --verbose --update cache download fetch http://nl.alpinelinux.org/alpine/edge/main/x86_64/APKINDEX.tar.gz fetch http://nl.alpinelinux.org/alpine/edge/community/x86_64/APKINDEX.tar.gz WARNING: This apk-tools is OLD! Some packages might not function properly. fetch http://nl.alpinelinux.org/alpine/edge/main/x86_64/mariadb-dev-10.1.28-r1.apk ERROR: mariadb-dev-10.1.28-r1: IO ERROR 2018-01-24 18:48:37 You check your free space? 2018-01-24 18:48:58 70.4G available 2018-01-24 18:49:15 WARNING: This apk-tools is OLD! 2018-01-24 18:49:22 maybe update your apk-tools first? 2018-01-24 18:50:12 I can say that clang definitely works on PPC and Boost works on everything I've tossed it at (including some embedded-ish stuff) 2018-01-24 18:50:38 I know LLVM (which Clang is a part of) supports ARM and MIPS as well but I haven't tried it 2018-01-24 18:50:54 Using tar with the downloaded file reports: gzip: crc error 2018-01-24 18:51:05 I expect the file on the remote server is actually bad 2018-01-24 18:55:15 Updating apk-tools didn't help. sudo apk --verbose --update cache download fetch http://nl.alpinelinux.org/alpine/edge/main/x86_64/mariadb-dev-10.1.28-r1.apk ERROR: mariadb-dev-10.1.28-r1: IO ERROR 2018-01-24 18:57:21 Where can I check to see the full list of APKBUILD functions? Is it possible to check the arch on the build() function? 2018-01-24 18:58:05 tcely: looks like the sha256sum for that package on nl is different from dl-2 and dl-3 (which both match), I think you are right. 2018-01-24 18:59:18 nm I see how they did it with the gcc APKBUILD 2018-01-24 19:09:43 When it compiles the packages? What machine can we assume it will do the cross compiling or packaging? Some packages do not have nodejs,go available per arch. 2018-01-24 19:10:07 I think I need to make a conditional makedepends per arch. 2018-01-24 19:13:11 <_ikke_> No cross-compilation happens 2018-01-24 19:13:21 <_ikke_> Dedicated builders 2018-01-24 19:14:54 In my situation for ycmd APKBUILD. It needs to compile with rust if available on x86_64 so if the arch is x86_64 it gets added to makedepends and the build() enables rust in the build script. 2018-01-24 19:15:13 build() function* 2018-01-24 19:40:55 nangel: nl does seem to be the odd source out. Unfortunately, that's the one travis seems to be using. https://pastebin.com/raw/UbEVjUJm 2018-01-24 19:41:35 Is there a specific reason to use nl instead of dl-cdn? 2018-01-24 20:03:26 what is wiki.intra.a.o that danieli / kaniini were discussing? 2018-01-24 20:03:33 haven't seen that before 2018-01-24 20:03:42 that's something clandmeter set up 2018-01-24 20:03:53 <_ikke_> tdtrask: Internel wiki you can access via vpn 2018-01-24 20:04:03 where'd you hear of wiki.inta? 2018-01-24 20:04:12 intra* 2018-01-24 20:04:15 danieli: you mentioned it in this channel 2018-01-24 20:04:21 not recently 2018-01-24 20:04:33 ACTION is catching up on backlog 2018-01-24 20:04:37 aha 2018-01-24 20:05:04 anyway, there is some non-public infrastructure 2018-01-24 20:05:09 ACTION tends to read a week's worth of IRC in a day because of lack of time :( 2018-01-24 20:05:15 <_ikke_> ACTION msg'd tdtrask the ip address 2018-01-24 20:05:26 _ikke_: the ip...? 2018-01-24 20:05:34 <_ikke_> of wiki.intra 2018-01-24 20:05:39 that won't help though 2018-01-24 20:05:45 not really 2018-01-24 20:05:55 without VPN access 2018-01-24 20:05:58 ^ 2018-01-24 20:06:02 <_ikke_> Ah, thought you had access 2018-01-24 20:06:05 lol 2018-01-24 20:06:23 ACTION is a second class developer :/ 2018-01-24 20:06:35 I'm gonna set up a little dnsmasq for resolving internal names 2018-01-24 20:06:39 <^7heo> nobody has access to anything, that's #alpine 2018-01-24 20:06:51 <_ikke_> danieli: challenge is to keep it up-to-date 2018-01-24 20:06:53 <^7heo> and who even knows what runs where? 2018-01-24 20:06:58 _ikke_: try me 2018-01-24 20:06:58 ACTION has push access to git, which is what he cares about :) 2018-01-24 20:07:07 ^7heo: not everyone needs access to everything 2018-01-24 20:07:10 <^7heo> tdtrask: yeah that's what people usually get 2018-01-24 20:07:13 <_ikke_> danieli: I didn't say it was impossible :-) 2018-01-24 20:07:48 _ikke_: I do a lot of recon security wise - I'm very used to handling huge amounts of fairly destructured data 2018-01-24 20:08:14 <_ikke_> Yeah, but the goal is to have it automated 2018-01-24 20:08:18 of course 2018-01-24 20:08:24 I'm gonna tinker with it when I finish some projects 2018-01-24 20:08:26 <^7heo> danieli: nah, true; but it would help developers to have access to documentation about what does waht. 2018-01-24 20:08:32 <^7heo> s/waht/what/ 2018-01-24 20:08:39 ^7heo: they don't need access to wiki.intra 2018-01-24 20:08:47 the relevant documentation is in the source, at docs.a.o and wiki.a.o 2018-01-24 20:08:51 and the mailing list 2018-01-24 20:09:05 <^7heo> The documentation has changed a LOT in the last months or it's terrible. 2018-01-24 20:09:10 <^7heo> I don't know, didn't look, didn't care. 2018-01-24 20:09:36 docs.a.o isn't too useful yet 2018-01-24 20:09:47 documentation really needs a shine-up 2018-01-24 20:10:03 ACTION wishes he had more time to help, sorry 2018-01-24 20:12:10 me as well, I'll have more time soon 2018-01-24 20:15:41 <^7heo> So as danieli kindly convinced me in /query 2018-01-24 20:15:45 <^7heo> I'll be leaving now. 2018-01-24 20:16:06 <^7heo> Turns out I'll probably be better off on BSD channels, people are not making a drama out of half finished sentences there. 2018-01-24 20:16:13 <^7heo> ACTION winks at kaniini  2018-01-24 20:16:22 <^7heo> ACTION winks at danieli too 2018-01-24 20:16:30 <^7heo> good luck with all that. 2018-01-24 20:16:39 <^7heo> I'm sure you'll be super successful (I'm not sarcastic) 2018-01-24 20:16:50 <^7heo> But I'm not sure you'll all like what you see when you look back at what you've done. 2018-01-24 20:31:06 well, that was random 2018-01-24 20:31:28 <_ikke_> yup 2018-01-24 20:33:34 what? 2018-01-24 20:34:28 <_ikke_> the whole explosion from ^7heo 2018-01-24 20:34:42 lol 2018-01-24 20:34:49 _ikke_: you saw query, right? 2018-01-24 20:34:53 <_ikke_> danieli: yea 2018-01-24 20:34:55 mm 2018-01-24 20:35:10 that's faaar from the first time 2018-01-24 20:35:16 <_ikke_> I know 2018-01-24 20:35:50 he consider(s|ed) me a friend so I guess me not being tolerant this time caused him to just /part 2018-01-24 20:35:51 oh well 2018-01-24 20:36:25 tolerant to what 2018-01-24 20:36:44 <_ikke_> kaniini: ^7heo pm'ed danieli 2018-01-24 20:36:47 yep 2018-01-24 20:37:03 logs ? 2018-01-24 20:37:31 query in a sec 2018-01-24 20:41:33 i see 2018-01-24 20:41:42 yep 2018-01-24 20:44:22 i will say this: he isn't wrong 2018-01-24 20:44:29 that is why i requested a full audit of *everything* on behalf of core 2018-01-24 20:44:43 about that part, yes, he's not wrong 2018-01-24 20:44:58 from what i gather, even you infra people do not even know what you have 2018-01-24 20:45:07 I have no idea, I'm fairly new, and I hardly have access to anything 2018-01-24 20:45:22 exactly 2018-01-24 20:45:34 VPN, one box, git access to infra/mirrors and wiki.intra.a.o 2018-01-24 20:45:41 I was not aware an audit was requested - are you looking for a security report, or just an inventory listing? 2018-01-24 20:46:10 <_ikke_> I think danieli and I can help at least mapping things 2018-01-24 20:46:11 a list of assets the project holds is a good starting point 2018-01-24 20:46:21 gotcha. I can do that part, given access 2018-01-24 20:46:34 I do a lot of devops and security work so :P 2018-01-24 20:46:58 there isn't any, uh, fragile infrastructure is there? there's no embedded devices that can't handle some light port scanning? 2018-01-24 20:47:03 this is something that need be sorted, so the project knows what it has to work with 2018-01-24 20:47:03 wait - #alpine-infra 2018-01-24 20:47:07 agreed 2018-01-24 20:47:31 if for example we decide to take an initiative that requires commitment of infrastructure 2018-01-24 20:47:41 we have no idea if the infra team can actually deliver on it 2018-01-24 20:47:51 which is really bad 2018-01-24 20:47:54 I had no idea that was even requested 2018-01-24 20:48:09 i think it hasn't been formally requested *yet* 2018-01-24 20:48:14 but it is something we have been thinking about doing 2018-01-24 20:48:28 because from the perspective outside infra 2018-01-24 20:48:30 you guys have some boxes here 2018-01-24 20:48:37 some boxes there 2018-01-24 20:48:42 how it all fits together is largely a mystery 2018-01-24 20:48:49 I don't think they do fit together too well, lol 2018-01-24 20:49:39 then there is the secret infra team wiki ;) 2018-01-24 20:49:57 which i mean yeah, i get why that is needed, but some of the information there is probably fine to declassify 2018-01-24 20:50:02 secret infra team wiki? 2018-01-24 20:50:06 wiki.intra? 2018-01-24 20:50:11 <_ikke_> yeah 2018-01-24 20:50:15 it's just some work in progress, not really a thing 2018-01-24 20:50:17 yes 2018-01-24 20:50:26 aha 2018-01-24 20:50:33 <_ikke_> I think most can go into docs.a.o 2018-01-24 20:50:39 anyway my point is 2018-01-24 20:50:46 the project needs more transparency in terms of infra 2018-01-24 20:51:10 agreed to both 2018-01-24 20:51:33 in a nutshell that is basically all 2018-01-24 20:57:46 For the arches list in gcc/APKBUILD there are more listed for the target triplets in the conditional than the number of Architecture categories in the package database. Does that mean the unmentioned are not supported or they belong to some general category? Is there a table that maps the triplet to the Architecture? 2018-01-24 21:32:21 I'm still confused about this CBUILD, CHOST, CTARGET. When the build tools run, what does CBUILD refer to? Does it refer to the environment holding the cross compiler/abuild or does it refer to the triplet within the cross compiled environment? 2018-01-24 22:20:11 the latter. see `abuild help` for more info 2018-01-24 23:28:39 nmeum: thanks for your work on cryptsetup 2018-01-24 23:53:37 the issue with mkinitfs on the builders is a bit tricky 2018-01-24 23:53:43 what happens is 2018-01-24 23:53:50 alpine-sdk depends on mkinitfs 2018-01-24 23:53:56 due to it is needed when building the iso image 2018-01-24 23:54:38 cryptsetup2 is built, but cannot be upgraded due to mkinitfs depend on the old 2018-01-24 23:55:14 we have two options 2018-01-24 23:56:03 1) we can uninstall alpine-sdk on the builders, and introduce a alpine-release-sdk which is a dummy package depending on everyting needed to create a release 2018-01-24 23:56:52 we could potentially also remove the rsync dependency and install/uninstall rsync when copying the built packages 2018-01-24 23:57:30 2) we create a migration package with cryptsetup1, which provides the old cryptsetup library 2018-01-24 23:58:02 from my experience alpine-sdk is not needed for building most packages, only build-base is 2018-01-24 23:58:07 which should satisfy the mkinitfs depends while cryptsetup2 is built 2018-01-24 23:58:33 ollieparanoid: exactly. thats why i thought that we could remove some buildtime deps 2018-01-24 23:58:53 the proper solution is to create the entire build env from scratch every build in an isolated container 2018-01-24 23:59:22 and only include build-base, not the full alpine-sdk 2018-01-25 00:01:03 hum 2018-01-25 00:01:40 adding cryptsetup1 will not work 2018-01-25 00:02:36 because there is nothing that says that cryptsetup1 should be built first 2018-01-25 00:04:05 i wonder if we should revert and add cryptsetup2, then we make all packages makedepend on cryptsetup2-dev instead 2018-01-25 00:04:31 once everything is built with cryptsetup2, we purge cryptsetup and rename cryptsetup2 2018-01-25 00:05:44 no, better only make mkinitfs depend on cryptsetup1 explicitly 2018-01-25 02:34:05 could someone do me a quick favor and check if this command segfaults alpine's ruby too ? 2018-01-25 02:34:09 $ irb 2018-01-25 02:34:09 irb(main):001:0> "foo" 2018-01-25 02:34:09 => "foo" 2018-01-25 02:34:09 irb(main):002:0> puts "xx" 2018-01-25 02:34:10 /lib/ruby/2.5.0/irb/ruby-token.rb:95: [BUG] Segmentation fault at 0x0000000000000010 2018-01-25 02:34:11 ruby 2.5.0p0 (2017-12-25 revision 61468) [x86_64-linux] 2018-01-25 02:34:48 i build it almost exactly as alpine, so i guess it should segfault there too... 2018-01-25 02:38:35 sh4rm4^bnc: did not seg 2018-01-25 02:39:20 dsabogal, thanks 2018-01-25 02:39:35 x86_64, right ? 2018-01-25 02:40:06 sh4rm4^bnc: yes 2018-01-25 03:02:21 Would it be necessary to prefix emacs packages with emacs- ? 2018-01-25 03:15:06 In Arch Linux they prefix it. In the Alpine documentation, it makes relevant for a few cases. 2018-01-25 13:26:52 looks like imagemagick testsuite still hangs on s390x 2018-01-25 13:34:45 anyone has an idea why gns3 fails due to this error? 2018-01-25 13:34:45 Fail update installation: Error relocating /usr/lib/python3.6/site-packages/PyQt5/QtCore.so: _ZN23QOperatingSystemVersion11AndroidOreoE: symbol not found 2018-01-25 13:34:46 Can't import Qt modules: Qt and/or PyQt is probably not installed correctly... 2018-01-25 13:35:02 py3-qt5 has been rebuilt against qt5 2018-01-25 13:35:27 smells like broken ABI 2018-01-25 13:35:43 TBK76: are you TBK on github? 2018-01-25 13:35:48 that-s right...but why? 2018-01-25 13:37:05 danieli, https://dpaste.de/0bAz 2018-01-25 13:37:19 oh wow 2018-01-25 13:37:31 it seems to me that py3-qt5 is broken agains qt5 2018-01-25 13:37:36 indeed seems like it 2018-01-25 13:37:59 wait hmm 2018-01-25 13:38:03 are you sure it isn't broken against python? 2018-01-25 13:38:06 look at the symbol names 2018-01-25 13:38:40 danieli, how can pyqt5 being broken against python itself? 2018-01-25 13:38:50 that's what amazes me 2018-01-25 13:39:08 lemme do some google-fu 2018-01-25 13:39:42 I have a feeling it has to do with python3-dev anyway 2018-01-25 13:40:17 danieli, maybe python3.6 is not supported? 2018-01-25 13:40:35 fcolista: what was the last python3-dev version pyqt successfully was built against? 2018-01-25 13:40:46 danieli, sure 2018-01-25 13:41:21 what? 2018-01-25 13:41:34 oh sorry, missread 2018-01-25 13:41:37 no worries 2018-01-25 13:41:37 python3.6/dev 2018-01-25 13:41:56 all right, sounds very strange to have a broken ABI between python3-dev minor version bumps 2018-01-25 13:42:05 it's edge btw 2018-01-25 13:42:14 still something that needs fixing 2018-01-25 13:42:42 https://www.riverbankcomputing.com/news/pyqt-5100 2018-01-25 13:42:51 pyqt 5.10 released 2 days ago 2018-01-25 13:43:02 that might be our issue 2018-01-25 13:43:18 qt5 version we have is not the latest one 2018-01-25 13:43:24 danieli: the broken ABI isn't in python 2018-01-25 13:43:27 it's in pyqt <-> qt 2018-01-25 13:43:29 we should have the problem on the other waty round 2018-01-25 13:43:34 judging by the symbol name 2018-01-25 13:43:42 Shiz, that's my guess to 2018-01-25 13:43:45 *too 2018-01-25 13:44:08 i think ollieparanoid wanted notification when we upgrade qt 2018-01-25 13:44:18 it's not a guess 2018-01-25 13:44:20 :P 2018-01-25 13:44:26 _ZN23QOperatingSystemVersion11AndroidOreoE = QOperatingSystemVersion::AndroidOreo 2018-01-25 13:44:30 yeah 2018-01-25 13:44:30 :P 2018-01-25 13:44:32 which is a Qt api 2018-01-25 13:44:45 pmOS needs to rebuild stuff when we upgrade qt 2018-01-25 13:45:07 Shiz, yes. Is python calling an API not available on qt then? 2018-01-25 13:45:10 ncopa: yeah, he should've been notified - does he want notifications even if upgrades happen in edge? 2018-01-25 13:45:28 yes 2018-01-25 13:45:43 looks like the pyqt version is targeted at a newer qt than we have 2018-01-25 13:45:46 so qt needs an upgrade 2018-01-25 13:45:50 ollieparanoid: ^^^ 2018-01-25 13:45:52 that-s wrong 2018-01-25 13:46:25 ABI does not change, and pyqt claims to be retro-compatible 2018-01-25 13:46:58 most symbols QtCore.so failed to find are python3-dev ones, only exception is the top one 2018-01-25 13:48:51 I haven't been messing a lot with pyqt since I did work on Qubes OS in 2014 or so 2018-01-25 13:52:09 fcolista: what is `ldd -r` saying? 2018-01-25 13:53:11 danieli, this: https://dpaste.de/QpHz 2018-01-25 14:20:51 sorry, had to take some time to fix a hijacked spotify account 2018-01-25 14:23:23 oh 2018-01-25 14:23:27 i only aw the broken Qt symbol 2018-01-25 14:23:33 yeah uhm 2018-01-25 14:23:38 that just looks like a broken link 2018-01-25 14:23:42 like 99.99% are from libpython 2018-01-25 14:23:42 yep 2018-01-25 14:23:42 not a changed python abi 2018-01-25 14:23:47 yeah, that'd be unlikely 2018-01-25 14:23:52 looks like it's missing a libpython DT_NEEDED entry 2018-01-25 14:24:13 oh wait it's ldd 2018-01-25 14:24:33 you can ignore the libpython symbols then 2018-01-25 14:24:39 in python libpython is already loaded into the process 2018-01-25 14:24:43 aah, right 2018-01-25 14:24:43 so those symols will resolve 2018-01-25 14:24:52 so it really is just the Qt symbol 2018-01-25 14:24:53 well yeah, then it's pyqt <=> qt 2018-01-25 14:24:55 14:46:58 danieli : most symbols QtCore.so failed to find are python3-dev ones, 2018-01-25 14:24:57 err 2018-01-25 14:24:59 14:46:25 @fcolista : ABI does not change, and pyqt claims to be retro-compatible 2018-01-25 14:25:01 clearly they're wrong then ;p 2018-01-25 14:25:05 yep 2018-01-25 14:25:12 or compiled against a newer Qt version than it's being linked against n the end 2018-01-25 14:25:15 (local build issue?) 2018-01-25 14:25:23 it's a version mismatch or broken ABI, I'm guessing the former 2018-01-25 14:31:13 ncopa: could having a small issue reporting program in alpine be an idea? i.e. run `report-bug` and it'll prompt for info, and create a bugs.a.o issue 2018-01-25 14:31:29 perhaps even attach certain log / config files 2018-01-25 14:31:40 (ref: #7494 ) 2018-01-25 14:32:35 having a full fledged GUI front end is a bit overkill, and writing said program in anything else than C/C++ would cause it to be heavyweight compared to for instance requiring python 2018-01-25 14:33:32 err 2018-01-25 14:34:02 s/compared to for instance/by/ 2018-01-25 14:42:31 danieli: i think we had one long time ago 2018-01-25 14:42:50 i think we had a sendbug.c at some point 2018-01-25 14:43:08 yeah, it'd have to be c or cpp to be small enough to ship with alpine 2018-01-25 14:43:21 what happened to it? do I close that issue as a wontfix? 2018-01-25 14:43:23 isnt that like 10y ago? 2018-01-25 14:43:34 could tell him he has to use his phone or another computer to report an issue 2018-01-25 14:43:44 most distributions have a simple little bug reporter though 2018-01-25 14:44:19 nobody used it 2018-01-25 14:44:26 i dont mind ressurect it 2018-01-25 14:44:31 the trickiest part is to configure the mailer 2018-01-25 14:44:40 do we want submit bugs via http or mail? 2018-01-25 14:44:48 or both 2018-01-25 14:44:55 I would use the redmine API and use HTTP 2018-01-25 14:44:58 but that's just me 2018-01-25 14:45:31 http is the most simple way 2018-01-25 14:45:35 indeed 2018-01-25 14:45:39 some ISPs block email ports 2018-01-25 14:45:52 nobody in their right mind would block 80 or 443 outwards on a normal network 2018-01-25 14:46:04 what happens if we replace redmine with something else? 2018-01-25 14:46:06 we deal with it then? 2018-01-25 14:46:08 yep 2018-01-25 14:46:10 <_ikke_> though you need to account for proxies 2018-01-25 14:46:35 could we make a proxy web app 2018-01-25 14:46:43 what? 2018-01-25 14:46:45 we could replace redmine first :) 2018-01-25 14:47:01 some networks require its users to proxy all web traffic through an internal proxy to reach the internet 2018-01-25 14:47:09 a proxy web app that recevies the report and forwards it to redmine or whatever bugtracker we use 2018-01-25 14:47:15 that's possible 2018-01-25 14:47:19 could write a little wrapper around it 2018-01-25 14:47:31 clandmeter: are we planning to get rid of redmine? 2018-01-25 14:47:50 ncopa: who created the alpine logo? can I talk to them, or do you happen to know what font is used? 2018-01-25 14:47:52 so we have a web api on receving the reports 2018-01-25 14:48:09 no plans yet afak. 2018-01-25 14:48:15 clandmeter: okay, all right 2018-01-25 14:48:29 ncopa: yes, a small web app functioning as a proxy of sorts to create bug reports on whatever bug tracker we use 2018-01-25 14:48:47 only the occasional complaints about redmine. 2018-01-25 14:48:47 i dont want end up in the situation where we want replace redmine but cannot because we have 10000 users that may need to send their bug reports to it 2018-01-25 14:49:08 it was a friend of mine who did the logo 2018-01-25 14:49:11 he is not involved in alpine 2018-01-25 14:49:19 yeah, it should probably be a separate project or category for reports from the bug report program in alpine 2018-01-25 14:49:21 <_ikke_> imo, redmine isn't that bad 2018-01-25 14:49:30 and we could invalidate the api key used by the proxy app at any time 2018-01-25 14:49:47 ncopa: all right, I wanted to modify it and create "alpine security" / "alpine infra" variants 2018-01-25 14:50:00 possibly just a different color palette (of the same shade) for security and infra 2018-01-25 14:50:14 i can ask him 2018-01-25 14:50:28 I have the vector graphics file so I really just need to know what font it is 2018-01-25 14:50:34 I could probably figure out myself but it's easier and more accurate to just ask 2018-01-25 14:50:44 if he has the original file, it'd be nice to have it available 2018-01-25 14:50:58 could even create a "press media kit" from it 2018-01-25 15:39:29 admin'd redmine for a while about ten years ago. yes, it's that bad. ;) 2018-01-25 15:39:55 (granted, most of my horror stories from that are due to ruby being a shitty platform...) 2018-01-25 15:40:46 yeah... also, slightly relevant: isrubyfastyet.com 2018-01-25 15:41:25 that page, lol. "I like to think that "fast" is, somewhat arbitrarily (...)" 2018-01-25 15:56:21 thoughts on #5720? 2018-01-25 15:57:01 ping clandmeter, ncopa, _ikke_, you guys are active and brainstorming is important 2018-01-25 15:57:40 are there any clients that does not redirect to https if connect to tcp 80 fails? 2018-01-25 15:58:18 <_ikke_> euhm, most? 2018-01-25 15:58:26 i am positive to it 2018-01-25 15:58:27 to https only 2018-01-25 15:58:28 could redirect http:// -> https:// 2018-01-25 15:58:44 i am also positive to it, it will make config mistakes and "oh shit" mistakes impossible 2018-01-25 15:59:04 I doubt we have any http clients needing access to any of our services that don't support https 2018-01-25 15:59:20 <_ikke_> afaik, you cannot just disable port 80 and expect every client to automatically redirect 2018-01-25 15:59:56 indeed, hence the reverse proxies I suggested could redirect to 443 2018-01-25 16:00:17 I have experience running haproxy, squid and (especially) nginx as reverse proxies 2018-01-25 16:00:52 most of our infra is behind nginx already 2018-01-25 16:00:53 btw, why are we talking infra in devel now? 2018-01-25 16:00:53 besides, we could lessen load on the actual web servers by caching static files on the reverse proxy 2018-01-25 16:01:25 oh, my apologies, I went off on a tangent - only meant to post here for attention from people not in #alpine-infra 2018-01-25 16:36:05 i wanna try out alpines's apk. i've built apk-tools, now what am i supposed to put into /etc/apk so it finds the repo ? 2018-01-25 16:37:41 <_ikke_> sh4rm4^bnc: /etc/apk/repositories 2018-01-25 16:37:48 <_ikke_> that's a plain text file with one repo per line 2018-01-25 16:38:09 ok, and what should be the content ? 2018-01-25 16:38:58 <_ikke_> One line with the url to a repo for example 2018-01-25 16:39:15 ncopa 2018-01-25 16:39:22 ok, and what is the url, _ikke_ ? ;) 2018-01-25 16:39:25 in my opinion, the way to do gcc8 2018-01-25 16:39:30 is to rename main/gcc to main/gcc6 first 2018-01-25 16:39:46 then we can introduce gcc8 with provides=gcc=8 2018-01-25 16:40:14 what is your thoughts 2018-01-25 16:40:16 <_ikke_> sh4rm4^bnc: http://dl-4.alpinelinux.org/alpine/edge/main/ would be an example 2018-01-25 16:40:20 because we need keep gcc6 for java 2018-01-25 16:40:23 thanks 2018-01-25 16:40:52 sh4rm4^bnc: hi, welcome! nice to have you here! 2018-01-25 16:41:36 hi ncopa... i'm frustrated with not getting ruby to not segfault, now i wanna try out if alpine's build work... 2018-01-25 16:41:56 sh4rm4^bnc: http://dl-cdn.alpinelinux.org/alpline/edge/main and http://dl-cdn.alpinelinux.org/alpine/edge/community 2018-01-25 16:42:01 dl-cdn is a fastly sponsored CDN 2018-01-25 16:42:08 and is pretty fast 2018-01-25 16:42:19 cool 2018-01-25 16:43:04 its super simple if you have docker.... 2018-01-25 16:43:13 <_ikke_> ncopa: I wanted to link to dl-cdn, but it redirects to dl-4 2018-01-25 16:43:33 heh 2018-01-25 16:43:55 :D 2018-01-25 16:44:14 i think i proposed that last time fastly broke 2018-01-25 16:45:13 lol 2018-01-25 16:45:30 ncopa so about that gcc idea 2018-01-25 16:45:48 ACTION wants to get gcc8 in so they can get riscv bootstrapped 2018-01-25 16:49:47 kaniini: so we create a gcc6 and make it possible to install it in parallel 2018-01-25 16:49:49 then we upgrade gcc to 8 2018-01-25 16:49:59 is that what we want to do? 2018-01-25 16:50:14 we need a gcc6 for bootstrapping java i suppose 2018-01-25 16:51:09 openat(3, "etc/apk/arch", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory) 2018-01-25 16:51:10 openat(3, "lib/apk/db/lock", O_RDWR|O_CREAT|O_CLOEXEC, 0600) = -1 ENOENT (No such file or directory) 2018-01-25 16:51:16 what is supposed to go there ? 2018-01-25 16:52:47 you need have the --initdb option on first run 2018-01-25 16:53:34 apk add --root $tmprootfs --initdb --update-cache alpine-base 2018-01-25 16:53:57 will install the 'alpine-base' package into the directory $tmprootfs 2018-01-25 16:54:18 great, ty 2018-01-25 16:54:29 kaniini: in the past we discussed here that maybe gcc could be split into gcc + gcc-java, so bumping gcc later would be less problematic. don't know whether anyone attempted such split, though. 2018-01-25 16:55:40 it would seem cleaner, but maybe harder to attain than going with full gcc6 2018-01-25 16:55:57 i was thinking of gcc6 instead of gcc-java, since we probably will need the entire gcc6 2018-01-25 16:56:40 i dont think you will be able to use gcc-java-6.4.0 with gcc 8 2018-01-25 16:58:00 if entire gcc6 is needed, then splitting is obviously out of question. I've never used gcc-java, so cannot comment whether it is or isn't possible 2018-01-25 17:01:34 i dont know either, but i would be surprised if you could use gcc-java 6 with gcc 8 2018-01-25 17:04:49 as long as gcj is standalonish enough (it would be nice if it was, no idea if it really is), it should be possible to coexist with newer gcc stuff. sure, gcj may reuse some common gcc stuff, so this needs to be available from gcc6, of course, but I suspect it should be still (much?) less than entire gcc, and depending on the ratio it could be worth pursuing or not 2018-01-25 17:10:12 ncopa, doing that i get ERROR: unsatisfiable constraints: alpine-base (missing): required by: world[alpine-base] 2018-01-25 17:13:15 did you add --update-cache? 2018-01-25 17:14:01 you proabably also need the repository: apk add --root $tmprootfs --initdb --update-cache --repository http://dl-cdn.alpinelinux.org/alpine/edge/main alpine-base 2018-01-25 17:14:21 apk will look for $tmprootfs/etc/apk/repositories 2018-01-25 17:14:34 yeah, i copied that there 2018-01-25 17:14:54 apk add --root /tmp/alpine --initdb --update-cache alpine-base <- thats the cmd i used 2018-01-25 17:15:18 which does not exist, so you need to add --repository option 2018-01-25 17:15:45 fetch http://dl-cdn.alpinelinux.org/alpine/edge/main/x86_64/APKINDEX.tar.gz 2018-01-25 17:15:45 ERROR: http://dl-cdn.alpinelinux.org/alpine/edge/main: UNTRUSTED signature 2018-01-25 17:16:17 right, the keys too 2018-01-25 17:16:22 :) 2018-01-25 17:16:30 how to get ? 2018-01-25 17:17:00 there is an option: --keys-dir 2018-01-25 17:18:12 https://nl.alpinelinux.org/alpine/edge/main/x86_64/alpine-keys-2.1-r1.apk 2018-01-25 17:18:19 and untar it 2018-01-25 17:18:26 its over https at least 2018-01-25 17:19:46 where they are stored typically ? 2018-01-25 17:20:12 sh4rm4^bnc: this script does similar what you are trying to do: https://github.com/lxc/lxc/blob/master/templates/lxc-alpine.in 2018-01-25 17:20:17 it includes checksums of the keys 2018-01-25 17:21:11 and it also verifies the signature of the apk.static 2018-01-25 17:21:27 they are typically stored in $root/etc/apk/keys 2018-01-25 17:32:34 yay, it works <3 2018-01-25 17:35:01 OK: 7 MiB in 16 packages 2018-01-25 17:35:02 now to install ruby there, i do just: apk add --root /tmp/alpine/ ruby-full ? 2018-01-25 17:42:44 yes, that should work 2018-01-25 17:42:44 you may need add the repos in /tmp/alpine/etc/apk/repositories 2018-01-25 17:42:45 oh, im stupid. We ship official alpine rootfs tarballs... 2018-01-25 17:42:46 you could have just untarred one of those.. 2018-01-25 17:43:04 no, its about being able to use apk tool 2018-01-25 17:43:07 if write a package to handle triggers and then later install another package without a trigger but alters that directory which the trigger monitors will it call the old trigger? 2018-01-25 17:43:51 yes 2018-01-25 17:44:40 koldbrutality: for example mkfontcache is the tool that generates font caches. we added a trigger that monitors font directories 2018-01-25 17:45:13 when you install any package with a font (in the monitored direcotry) the fontcache trigger is run 2018-01-25 17:45:35 i'm curious if i can make apk coexist with butch on sabotage. i'm getting tired of maintaining an entire distro on my own... 2018-01-25 17:46:15 sh4rm4^bnc: i feel you... 2018-01-25 17:47:39 butch is implemented in shell/awk nowdays? 2018-01-25 17:47:43 yep 2018-01-25 17:48:16 awk is used in a background task so i can use its hashmap functionality, which sh lacks 2018-01-25 17:48:29 otherwise doing dependency scan would be too slow 2018-01-25 17:48:52 i think i do dependency scanning in awk too in abuild 2018-01-25 17:49:42 does butch generate a binary package of some sort? 2018-01-25 17:49:50 or are you supposed to build and install from source 2018-01-25 17:49:58 like gentoo 2018-01-25 17:50:05 not directly, but it can make a pack from an installed dir 2018-01-25 17:50:24 every package lives in its own dir in /opt , so its easy 2018-01-25 17:50:33 but yeah, usually its all from source 2018-01-25 17:50:57 i only use the pack feature to transfer huge stuff like firefox from a fast build box 2018-01-25 17:51:09 and butch installs precompled packs? 2018-01-25 17:51:28 understand 2018-01-25 17:51:32 butch unpack can do that. but by default, its about compiling source 2018-01-25 17:51:49 precompiled binaries are a second hand citizen 2018-01-25 17:51:50 like on gentoo 2018-01-25 17:52:12 yes 2018-01-25 17:52:17 second class citizen, you mean? :') 2018-01-25 17:52:21 so, with alpine, apk only manages the precompiled packages 2018-01-25 17:52:25 and we use abuild to compile the sources 2018-01-25 17:52:31 ah, i see 2018-01-25 17:53:07 second class* yes :) 2018-01-25 17:53:20 i suppose you could make butch generate .apk binaries too 2018-01-25 17:53:31 but it wouldnt really solve your maintenance problem 2018-01-25 17:53:40 yeah, quite possibly 2018-01-25 17:54:13 tbh, im not 100% happy with the APKBUILD format 2018-01-25 17:54:25 but i dont have a good solution for it either 2018-01-25 17:54:36 what's the issue ? 2018-01-25 17:54:55 various 2018-01-25 17:55:02 one issue is that its a bit hackish atm i think 2018-01-25 17:55:17 has grown organically, not well designed 2018-01-25 17:57:21 however well you design a package manager, there will always be unexpected surprises 2018-01-25 17:57:25 but i think the major issue is that the posix shell langugage does not support nested data structs 2018-01-25 17:57:25 like hashes.... 2018-01-25 17:57:26 i'd like to specify dependencies of subpackages in a nice way 2018-01-25 17:57:27 for example, we split package into a main runtime, and a -dev runtime 2018-01-25 17:57:28 and sometimes we split the shared libraries into a -libs subpackage 2018-01-25 17:57:28 when we define the subpackage, we need to set some metadata like description, dependencies 2018-01-25 17:57:53 some subpackage are noarch (only scripts, not arch dependent binaries) 2018-01-25 17:58:04 etc 2018-01-25 17:58:06 but major issue is dependencies 2018-01-25 17:58:10 i see 2018-01-25 17:58:52 for example, a C app may ship a small runtime tool written in perl 2018-01-25 17:58:59 perl is not needed for the main app 2018-01-25 17:59:11 perl is not needed build time 2018-01-25 17:59:29 and we dont want pull in perl as dependency unles needed 2018-01-25 17:59:31 kaniini: I have had the XPTI patches for 3.5, 3.6 and 3.7 running on there different dom0s since opening the PRs, seems solid. 2018-01-25 17:59:47 so we move the tool into its own subpackage, and make it depend on perl 2018-01-25 18:00:33 the current way we do that is that we override the depends= variable in the shell function that splits the package 2018-01-25 18:01:09 that means that we dont know about the perl dependency til after we have compiled the package and are about to split it 2018-01-25 18:01:23 that is a problem because we need to build perl before 2018-01-25 18:01:28 i see. that would be something like git-svn or git-add-patch 2018-01-25 18:01:49 correct 2018-01-25 18:02:16 sh4rm4^bnc: what is current status of your netbsd ncurses port 2018-01-25 18:02:42 it works well, i'm using it exclusively since < 1 yr 2018-01-25 18:03:04 the only real feature it lacks is mouse support 2018-01-25 18:03:20 though weirdly through some magic, in mc mouse works anyway 2018-01-25 18:03:21 i'd like to replace gnu ncurses 2018-01-25 18:04:09 yes, gnu ncurses maintenance situation is really bad 2018-01-25 18:04:19 yeah, that could cut some KBs in tiny rootfs 2018-01-25 18:04:49 oh ? i thought ncurses is dickey's darling 2018-01-25 18:07:36 s// btw 2018-01-25 18:07:57 imo the biggest advantage of netbsd curses is that you can make quite small, self-contained static binaries 2018-01-25 18:08:53 which is one of musl's strong qualities too 2018-01-25 18:09:16 we prefer shared in alpine, for maintenance reasons 2018-01-25 18:10:57 sh4rm4^bnc: he hasn't shipped any real release in a long time and snapshots get deleted and things 2018-01-25 18:11:00 it is a mess 2018-01-25 18:11:23 i ee 2018-01-25 18:11:28 *i see 2018-01-25 18:12:54 sh4rm4^bnc: i am open to suggestions what we can do to help sabotage 2018-01-25 18:13:10 we have a pretty good relation with void linux too 2018-01-25 18:13:17 well he already uses pkgconf (: 2018-01-25 18:13:27 :) 2018-01-25 18:17:15 hm 2018-01-25 18:17:16 we should use netbsd-curses 2018-01-25 18:17:21 i wonder how many users we have that needs mouse support in console 2018-01-25 18:17:33 ACTION raises hand 2018-01-25 18:17:51 need not really but it would confuse me if its gone 2018-01-25 18:18:07 mouse support in console, lol 2018-01-25 18:18:19 mitchy: what do you use it for? 2018-01-25 18:18:28 ncopa: emacs :) 2018-01-25 18:18:51 only emacs? 2018-01-25 18:18:52 the one windowing manager that works in gui and console mode (tm) 2018-01-25 18:19:02 well weechat and tmux too 2018-01-25 18:19:13 im thinking of things like `mc` 2018-01-25 18:19:17 tmux? 2018-01-25 18:19:18 hum 2018-01-25 18:19:18 i don't need it, but when switching tabs its super nice 2018-01-25 18:19:32 :set mouse on 2018-01-25 18:19:40 how do you switch tabs in tmux with mouse? 2018-01-25 18:19:46 click them, I believe 2018-01-25 18:19:52 yep click the name 2018-01-25 18:20:08 I haven't used tmux much with mouse, did for a bit like half a year ago 2018-01-25 18:20:12 ncopa: in tmux you can also resize tabs with the mouse 2018-01-25 18:20:16 mm 2018-01-25 18:20:19 and scroll 2018-01-25 18:20:25 or rather panes 2018-01-25 18:20:33 what terminal? i use xfce-terminal and it does not seem to work there 2018-01-25 18:20:40 yep, it comes in more handy than you'd think 2018-01-25 18:20:45 ah I use konsole 2018-01-25 18:21:11 i think vim has some mouse support but it confuses me :) 2018-01-25 18:22:05 well, as i said, mc have some magic built-in that makes mouse work anyway 2018-01-25 18:22:26 but not tmux and emacs? 2018-01-25 18:22:39 i didnt port emacs yet... 2018-01-25 18:23:02 i dunno who ported it to alpine 2018-01-25 18:23:11 but nope, mc is the only app i'm aware of that supports mouse 2018-01-25 18:23:12 does vim and neovim need porting as well? 2018-01-25 18:23:32 to the new ncurses 2018-01-25 18:23:40 emacs is using highly non portable tricks 2018-01-25 18:23:51 so the porting effort is about getting it to run on musl 2018-01-25 18:23:59 if we switch we'd try use it for everything 2018-01-25 18:24:00 ah 2018-01-25 18:24:24 yeah, emacs does (did?) some nasty stuff 2018-01-25 18:24:29 btw, adding mouse support might not be a huge effort 2018-01-25 18:24:46 like unexec or whatever that glibc hack was 2018-01-25 18:24:54 when it was added to termbox, iirc it was just like 150 lines 2018-01-25 18:25:13 yeah, they dump the current memory contents to harddisk 2018-01-25 18:25:26 then load the memory dump on startup, so it starts in a reasonable time 2018-01-25 18:25:27 anyway, don't take my usage of terminal stuff with the mouse as anything more than the mouse is useful even in the console 2018-01-25 18:25:40 it got ripped out of glibc, emacs was the only user 2018-01-25 18:25:43 seems like we have atleast a handful testers in case you'd like to add mouse support 2018-01-25 18:26:01 lol 2018-01-25 18:26:06 :> 2018-01-25 18:26:57 so they fixed emacs? 2018-01-25 18:27:18 https://lwn.net/Articles/673724/ 2018-01-25 18:27:53 my impression when i looked at it was: emacs seems like a nice OS. it only lacks a decent text editor... :) 2018-01-25 18:28:27 heh, just enable viper mode or one of the other 4 or more vi emulators if you want that :) 2018-01-25 18:35:26 hmm 2018-01-25 18:35:43 sh4rm4^bnc: why netbsd curses libs installed as foo.so instead of foo.so.0 or whatever 2018-01-25 18:45:15 the .so's also do not have SONAME set 2018-01-25 18:45:17 hmm 2018-01-25 19:14:35 is it xtables-addons-vanilla or linux-vanilla-dev that needs perl in depends? http://build.alpinelinux.org/buildlogs/build-edge-s390x/main/xtables-addons-vanilla/xtables-addons-vanilla-4.14.15-r0.log 2018-01-25 19:15:09 i think it's the headers 2018-01-25 19:15:55 it does not happen on the other arches 2018-01-25 19:16:01 only s390x 2018-01-25 19:22:41 kaniini, it didn't appear necessary to me 2018-01-25 19:34:45 is there a way to run the apk triggers for deleted files? 2018-01-25 19:35:25 or packages 2018-01-25 19:37:31 sh4rm4^bnc: in general, it is good to have proper SONAME and SOVERSION :p 2018-01-25 19:37:35 there are people around here who will be like 2018-01-25 19:37:36 omfg 2018-01-25 19:37:38 no SOVERSION 2018-01-25 19:37:46 no SONAME 2018-01-25 19:37:47 what is this 2018-01-25 19:37:59 so, are you cool with a patch to add that ;) 2018-01-25 19:38:12 well, i'm not a friend of cargoculting 2018-01-25 19:38:23 if there's a good reason to add it, sure 2018-01-25 19:38:40 well our package manager uses the SONAMEs 2018-01-25 19:38:54 and if there is ever an ABI break 2018-01-25 19:39:02 then it's good to have proper SOVERSION anyway 2018-01-25 19:39:09 ok, then go ahead 2018-01-25 19:58:43 huh, the perl script is not really needed 2018-01-25 19:58:50 for xtables-addons 2018-01-25 22:08:16 I'm going to make a pull request with 13 new packages. Is that okay. It has a dependency chain which some packages are not found in aports. 2018-01-25 22:09:00 Or, I should make my own GitHub repo for view before I make the pull? 2018-01-25 22:09:06 review* 2018-01-25 22:10:08 The fewer packages per PR you can manage, the better, but one commit per package, please. 2018-01-25 22:11:06 I am going to upload: company-mode, dash-el, deferred, emacs-site-start, f-el, flycheck, popup-el, request-el, rust-src, s-el, ycmd, youcompleteme 2018-01-25 22:12:17 I still have questions about the trigger system because the emacs-site-start seems to not to update for deleted packages. 2018-01-25 22:13:13 I might need to put the refresh plugins in post-deinstall 2018-01-25 22:15:38 /usr/share/emacs/site-lisp/site-start.el gets updated for the trigger in emacs-site-start when new emacs site package is detected which works fine with apk add but not apk del. 2018-01-25 22:27:24 nm the trigger seems to work 2018-01-25 22:37:48 Can I just make a pull request with just youcompleteme, ycmd. When ycmd gets accepted, I could finish with emacs-ycmd (main package), and the dependencies company-mode, dash-el, deferred, emacs-site-start, f-el, flycheck, popup-el, request-el. This way it could pass the CI tests. 2018-01-25 22:44:10 nm I will just do it one by one 2018-01-25 23:14:15 yes that makes a lot of sense :) 2018-01-25 23:29:38 For the licensing field in the APKBUILD, do I need to put the third party license as well? Because it would give the wrong impression for the primary license. 2018-01-25 23:30:20 The licenses of JavaScript modules are MIT, but the main project is GPL-3. 2018-01-25 23:31:22 Is it implied inclusive? or mutually exclusive? 2018-01-25 23:53:49 so many licenses lol 2018-01-26 08:51:19 Hi is possible here ask anyone about APK package development? I must create php package in versions 5.4, 5.5 that are not actually supported. I copied APKBUILD from Alpine 2.6 where is possible and edited it. 2018-01-26 08:52:16 There is last version 5.4.40, but last version on php.net is 5.4.45. When I would like to build 5.4.45 version, I have problem with build. 2018-01-26 08:53:30 Is there anyone who build php 5.4 under version 3.7? Is it possible? 2018-01-26 08:59:27 We need to see the debug log 2018-01-26 08:59:42 or the context of the error 2018-01-26 10:34:16 what koldbrutality said 2018-01-26 12:06:42 what Shiz said 2018-01-26 14:06:55 what danieli said 2018-01-26 14:07:13 i believe it is possible to build php 5.4 under version 3.7 2018-01-26 14:07:30 everything is possible. the impossible just takes longer time 2018-01-26 14:07:42 should be possible, yes 2018-01-26 14:07:55 with the right mindset, everything is :P 2018-01-26 14:08:32 I think building it under 3.7 sounds like a better idea 2018-01-26 16:10:28 koldbrutality: sorry for overlooking your license question. you should mention all licenses that stuff in your package uses, ideally starting from the most commonly used one. 2018-01-26 16:11:00 we're using SPDX identifiers for that: https://spdx.org/licenses/ 2018-01-26 16:12:10 it's not fixed for most packages yet, but we'll get there eventually 2018-01-26 16:12:20 yeah, it's gonna be a tedious process 2018-01-26 16:13:39 what if the package uses nodejs? do we push that to the front? 2018-01-26 16:13:47 node js packages? 2018-01-26 16:14:00 if it uses nodejs, it'll depend on the package for nodejs, with its own license set 2018-01-26 16:14:13 you only specify licenses of the code that is in your package 2018-01-26 16:14:32 ^ 2018-01-26 16:14:37 npm and cargo pulls a lot of stuff 2018-01-26 16:14:48 elaborate? 2018-01-26 16:15:43 the primary package states it is gpl-3.0-or-newer, but you said the most used one but they are many packages. many of the tiny third party packages use MIT and ISC. 2018-01-26 16:17:01 in alpine, let's say you have a py3-mywebapp that depends on python3 and the python package requests - you don't use pip to download requests, you make your package depend on py3-requests 2018-01-26 16:17:52 in each package we're specifying only licenses of the code that is shipped (be it script or compiled version) within this packages. if there are deps, those dep packages should specify their license they use and so on. 2018-01-26 16:19:23 ...but the build script pulls in additional packages. those code gets compiled in and shipped with the package. 2018-01-26 16:19:29 you don't use npm to pull dependencies 2018-01-26 16:19:34 you depend on existing alpine packages 2018-01-26 16:20:25 well this software does it or else aports will be flooded with cargo and npm packages 2018-01-26 16:20:56 if you need `colorama` from pypi, you create a py(2|3)-colorama package 2018-01-26 16:21:16 python it is easy, the dependencies are light. 2018-01-26 16:21:29 but rust and node it is more of a nightmare 2018-01-26 16:21:46 heaviest objects in the universe: #1. node_modules 2018-01-26 16:25:05 well, I think they're not as heavy as stars yet, so you're exaggerating a bit, but let's leave whether node is nice, heavy or not, outside of this discussion 2018-01-26 16:26:40 I'm generally of the opinion, at this point, that software supplied in language-specific repositories shouldn't be packaged by Alpine unless desktop or non-webapp software actually needs it. Doing otherwise creates a maintenance burden, and webapps (especially on r 2018-01-26 16:26:42 it's a meme, but yeah, let's leave that out 2018-01-26 16:27:08 Aerdan[m]: other distributions package a lot of language-specific packages for a reason 2018-01-26 16:27:11 and webapps (especially on Ruby) depend sometimes on differing versions of the same software that can't coexist. 2018-01-26 16:27:41 Aerdan[m]: does matrix not support message splitting? 2018-01-26 16:27:57 it seemed like you manually repeated part that didn't fit earlier 2018-01-26 16:28:08 I hit enter too early. Matrix cna't fix that. ;) 2018-01-26 16:28:17 ah, ok 2018-01-26 17:14:57 Is there a way to add a custom license like NOKIA-1.1-LGPL-2.1-exception? 2018-01-26 17:15:19 <_ikke_> just put custom and add the license in the license path 2018-01-26 17:15:42 what you mean license path? 2018-01-26 17:15:55 is there a licence folder like in gentoo portage? 2018-01-26 17:16:33 or we just put the license in the same folder as the package? 2018-01-26 17:16:45 <_ikke_> https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#license 2018-01-26 17:18:32 I see. I've been doing something similar. Are we permitted to put custom identifiers in license=? 2018-01-26 17:19:03 <_ikke_> koldbrutality: They do try to standardize them 2018-01-26 17:19:15 <_ikke_> https://spdx.org/licenses/ 2018-01-26 17:19:16 preferably use SPDX compatible license strings 2018-01-26 17:19:16 Or it only works with the list by SPDX 2018-01-26 17:19:25 it'll strictly work with anything but we want to normalize it 2018-01-26 17:19:27 There is not LGPL with exception 2018-01-26 17:19:44 <_ikke_> koldbrutality: then choose custom 2018-01-26 17:20:16 could write a comment in the APKBUILD for anyone actually looking it up 2018-01-26 17:20:44 oh you meant create a custom package or put it in the doc folder... i've been putting licenses in the doc package. 2018-01-26 17:31:38 <_ikke_> koldbrutality: license="custom", then install -Dm644 COPYING "$pkgdir"/usr/share/licenses/$pkgname/COPYING 2018-01-26 17:31:46 <_ikke_> where COPYING is your license file 2018-01-26 17:32:02 thx 2018-01-26 17:32:32 On Gentoo you can put custom identifiers 2018-01-26 17:42:15 i think some of us does not think "custom" is a valid license 2018-01-26 17:43:36 <_ikke_> ncopa: What should they specify then? 2018-01-26 17:45:00 i think jirutka has pretty strong opinion there, but not sure how to specify it properly 2018-01-26 17:46:14 <_ikke_> Well, SPDX is a good reference, but it's not all exhaustive 2018-01-26 17:46:15 i suppose you could write like "NOKIA-1.1-LGPL-2.1-exception" instead of "custom" and then add the LICENSE or COPYING file 2018-01-26 17:47:06 jirutka: how do you suggest to specify the license "NOKIA-1.1-LGPL-2.1-exception"? 2018-01-26 17:50:05 sorry guys, I was into different matters, so couldn't take part in this discussion 2018-01-26 17:50:20 it's already supported by SPDX 2018-01-26 17:50:44 LGPL-2.1-only WITH Nokia-Qt-exception-1.1 2018-01-26 17:50:59 https://spdx.org/licenses/exceptions-index.html 2018-01-26 17:54:15 They should add that url in the Alpine documentation 2018-01-26 17:55:24 <_ikke_> The SPDX list is in there 2018-01-26 17:55:35 <_ikke_> And that links to exceptions 2018-01-26 17:55:36 The decision to use SPDX is a recent development, I thought. 2018-01-26 17:55:40 <_ikke_> it is 2018-01-26 17:55:59 it's not that recent, it happened in June 2017 2018-01-26 17:56:29 <_ikke_> that's fairly recent 2018-01-26 17:56:30 but wasn't communicated properly to broader audience. 2018-01-26 17:58:51 currently licenses are listed with space as separator, and it has to be fixed. I think that comma is the best choice. 2018-01-26 18:08:07 I'll send some mail regarding how I would like to approach fixing this "license mess", as clandmeter nicely put it. my last mail didn't get public response, but we talked a bit with clandmeter in PMs since then, so I'll write possible ways of moving forward in that upcoming mail and depending on what solution will be most favorable, I'll simply start doing it. 2018-01-26 18:08:37 I hope I'll manage to write it this weekend. 2018-01-26 18:17:11 i think jirutka started to work on it too 2018-01-26 18:18:14 <_ikke_> Any concrete plans btw on the membership anouncements? 2018-01-26 18:19:53 we are working on it 2018-01-26 18:20:05 <_ikke_> ok 2018-01-26 18:20:59 btw.. what do you mean with "membership announcements"? 2018-01-26 18:21:29 we are working on a formal process on how you get membership (and how you lose it) 2018-01-26 18:21:43 basically, it is documenting how it currently works, more or less 2018-01-26 18:22:10 <_ikke_> ncopa: William Pitcock sent anouncements on alpine-devel 2018-01-26 18:22:31 ncopa: there are some license fixes in aports already, true, some of them are jirutka's, but this work is, AFAICT as casual observer, not coordinated in any way and you cannot easily tell whether APKBUILD license field was already reviewed or not. I think it needs to be fixed first, before this work continues (I wouldn't even start it before that, to be honest, as we have too many pkgs). 2018-01-26 18:23:02 _ikke_: ah :D 2018-01-26 18:24:11 ncopa: yep, 7 days ago at 9:18 to alpine-devel 2018-01-26 18:24:30 pm, that uis 2018-01-26 18:24:31 is* 2018-01-26 18:25:05 przemoc: ok. I think jirutka should coordinate the license cleanup 2018-01-26 18:25:09 and while jirutka also did automated regexp-based changes, which are appreciated, they are not enough, as packages may have (and definitely had in the past) wrong license field. 2018-01-26 18:25:45 przemoc: the license field can be reeeeally strange in some packages 2018-01-26 18:25:50 it's just a mess 2018-01-26 18:26:12 I will send mail tomorrow or on Sunday with some proposals 2018-01-26 18:27:17 jirutka: can you take reponsibility for the "clean up licenses" project? 2018-01-26 18:28:17 przemoc: please send an email! 2018-01-26 18:30:56 Do we need to add developer licenses in the license= field? 2018-01-26 18:31:16 Only relevant to the build process but not the distribution to end users? 2018-01-26 18:32:14 The only reason I would put the dev licenses there is for a future APKBUILD maintainer. 2018-01-26 18:34:36 license= field is for the built runtime 2018-01-26 18:34:39 license= is for end user, and license_build= for abuild user, and license_dev= for dev packages 2018-01-26 18:34:53 ok 2018-01-26 18:35:35 ugh.. sounds complicated 2018-01-26 18:35:44 licence_dev may make sense 2018-01-26 18:36:02 the thing is it uses BSD for build scripts but the final distributed product doesn't use it 2018-01-26 18:36:06 if licese of -dev package is different than the runtime 2018-01-26 18:38:23 is $pkgname-dev-doc standard? 2018-01-26 18:38:56 or i put the developer docs in the $pkgname-doc instead? 2018-01-26 18:48:28 I'd put the developer docs in -doc 2018-01-26 18:48:40 $pkgname-dev-doc is not standard 2018-01-26 21:57:08 Is this form of this license https://github.com/eclipse/m2e-core/blob/master/m2e-maven-runtime/org.eclipse.m2e.archetype.common/src/main/resources/about_files/jdom-LICENSE.txt found in SPDX list? 2018-01-26 22:02:17 They said a modified Apache... It looks like Apache 1.1 2018-01-27 01:19:25 could an alpine developer look into these and merge them if they look good? https://github.com/alpinelinux/aports/pull/3045 https://github.com/alpinelinux/aports/pull/3048 https://github.com/alpinelinux/aports/pull/3132 i've personally used and tested them in production in the last two weeks without issues 2018-01-27 01:37:50 Fusl: 18 days... :-/ 2018-01-27 01:37:53 sorry it is taking time 2018-01-27 01:48:38 Fusl: i commented on a few nitpicks 2018-01-27 01:56:38 Fusl: hum.. seems like it failed to build on armhf and x86: http://build.alpinelinux.org/buildlogs/build-edge-armhf/testing/mbuffer/mbuffer-20171011-r0.log 2018-01-27 01:59:32 i think its a race in the make install 2018-01-27 02:07:22 I added a few fixes 2018-01-27 02:13:27 ncopa: thank you 2018-01-27 02:30:30 i commented on the others. imho they are good enough. ping me if they are not merged tomorrow. need to sleep now. thanks! 2018-01-27 02:51:17 I made a pull request for keepassxc. 2018-01-27 10:40:04 what and why is https://git.alpinelinux.org/cgit/testing? 2018-01-27 12:40:44 what user agent does apk use when contacting mirrors? 2018-01-27 12:56:10 libfetch/2.0 2018-01-27 12:56:17 as of 3.7 2018-01-27 12:56:33 apk-tools 2.8.2 2018-01-27 12:58:25 good to know 2018-01-27 12:58:27 thank you 2018-01-27 13:02:19 defined here: https://git.alpinelinux.org/cgit/apk-tools/tree/libfetch/fetch.h#n39, used here: https://git.alpinelinux.org/cgit/apk-tools/tree/libfetch/http.c#n963 2018-01-27 13:03:32 figured as much - wasn't sure where to look, didn't care to grep for it 2018-01-27 14:30:47 kaniini: trying to get a octeon kernel running, but looking at the package the files in /boot contains the version string, e.g. vmlinux-4.14.14-0-vanilla 2018-01-27 14:31:10 this is not the case in packages for the other archs, so I'm wondering where I've went wrong 2018-01-27 15:58:47 Lochnair: it seems that the octeon doesn't support CONFIG_SYS_SUPPORTS_ZBOOT=y (which determines whether a vmlinuz is built or not). we could use a vmlinux, but certain scripts (e.g. update-kernel/mkinitfs) assumes that there is a vmlinuz 2018-01-27 16:27:33 is there any documentation on getting abuild-rootbld to work? 2018-01-27 16:59:51 dsabogal: looks like I need to get my eyes checked. I'll probably just create a profile for the Octeon-based EdgeRouters and the necessary scripts then 2018-01-27 17:00:07 I'd prefer to create an image that can be flashed easily from the stock FW 2018-01-27 17:07:40 What does https://pastebin.com/X7EDAwR2 mean? 2018-01-27 17:35:40 koldbrutality: looks like one of the metadata entries is excessively long 2018-01-27 17:35:45 what does the APKBUILD look like? 2018-01-27 17:37:52 https://pastebin.com/AQ3ARGk8 2018-01-27 17:38:53 looks like the license field may be excessively long 2018-01-27 17:39:53 that's a metric ton of licenses 2018-01-27 17:40:07 Those get shipped out 2018-01-27 17:40:22 I could try to sort then run uniq 2018-01-27 17:40:33 yeah 2018-01-27 17:40:35 also 2018-01-27 17:41:08 i'm not sure if you plan upstreaming this but general packaging practices within alpine would be to replace vendor'd packages by system variants 2018-01-27 17:41:16 so that would cut out some of that issue anyway 2018-01-27 17:42:17 It's a complicated answer. Like I said nodejs and cargo rust have more dependencies than your usually python and c dependencies. 2018-01-27 17:43:11 Same with this java more dependencies per project than usual so more licenses pulled. 2018-01-27 17:45:01 If we unbundle then you will see over 500+ additional tiny packages flooding aports. 2018-01-27 17:50:04 this sure sounds like the webapp situation I talked about a day or so ago. 2018-01-27 17:50:36 You should see my gentoo-overlay I unbundled it all. Maintenance nightmare. 2018-01-27 17:51:41 alpine the package system doesn't like slot system so it doesn't make sense. 2018-01-27 17:51:52 hurray for micro libraries! 2018-01-27 17:52:10 why not have a package for a single line of code? 2018-01-27 17:54:27 it shouldn't be a problem to have many packages, and there is some probability that other new packages will be able to use them too, as there may be other requests for node-ish or rush-ish packages in future. I am not into node or rust, so dunno what is the track record of node-ish or rust-ish packages so far in AL, but as python or lua have their modules packaged by distro, I'm not sure AL should 2018-01-27 17:54:33 play favoritism for some kind of ecosystems and disregard others, and if it will, it needs good rationale that is publicly documented. 2018-01-27 17:55:33 nodejs people's penchant for incorporating trivial libraries in lieu of just incorporating the trivial code itself is nothing if not absurd. 2018-01-27 17:57:20 przemoc: I can't speak for rust (as my opinions on rust are largely unprintable), but nodejs people have an unfortunate tendency to use hundreds of libraries, which means packaging nodejs software is a maintenance nightmare. webapps in other languages aren't that much better, though at least Python-based and Perl-based stuff only use a dozen or so directly. 2018-01-27 17:59:43 nodejs, python, and perl at least have ways to install deps locally to webapps, though (perl via local::lib, python via virtualenv, and nodejs just does node_modules by default). 2018-01-27 17:59:53 fortunately low-code will save us all from libraries 2018-01-27 18:01:47 I heard about librarizing oneliners, but also some horror stories about a lot of software not building anymore because someone pulled out his dozens packages from repo that half of world depended on. and the latter is kind of another reason, why distro should avoid depending on npms, cargos and other ecosystem-based package managers. I also understand it can be painful to package all that stuff, 2018-01-27 18:01:54 but AL at least doesn't prematurely package whole world - only the stuff people request or maintainerw want to be packaged. 2018-01-27 18:02:58 s/nerw/ners/ 2018-01-27 18:05:02 people should just stop using libraries for everything 2018-01-27 18:05:16 there wouldn't be any problem with a 13 line package being pulled if no one used it 2018-01-27 18:05:24 well, *that* horror story with someone pulling their code was mostly because a lot of people decided that depending on a library whose sole purpose was left-padding text was a good idea. instead of, y'know, writing the trivial function themselves like a sane person would. 2018-01-27 18:05:41 there was other stuff to it as well 2018-01-27 18:05:57 such as npm being crap and just handing over a package to another entity for no good reason 2018-01-27 18:06:20 it's just a shit package manager and a shit organization 2018-01-27 18:06:54 that was the reason the guy pulled his packages 2018-01-27 18:08:19 his code wasn't worth much regardless, so while it was a bit of a dick move for them to forcibly hand over the package name to the company, there wasn't much value lost. 2018-01-27 18:08:28 the real problem with node is that there's no namespacing. 2018-01-27 18:14:08 real problem with node is that you can't rely it for anything 2018-01-27 18:14:17 they lost all credibility 2018-01-27 18:16:15 anyway, I think that bundling lot of stuff in one package is short-sighted workaround, it's perfectly fine for some personal repos, not necessarily for official ones. what we most likely need is a way to lessen the burden of creating aports for those thousands of packages coming from node, rust and whatever will appear in future, and updating them too. 2018-01-27 18:16:55 what needs to happen is hand picking only the very best stuff 2018-01-27 18:17:01 and disregarding everything else 2018-01-27 18:17:23 it's too strict, because there may be great and valuable stuff that depends on lib you disregarded earlier 2018-01-27 18:18:01 I really don't think we should be packaging hundreds of nodejs libraries just for apps. the ecosystem is horrible from a distribution standpoint. 2018-01-27 18:18:03 something great and valuable that depends on a node or rust lib? 2018-01-27 18:18:15 you can make that argument when such software exists 2018-01-27 18:18:21 but that's currently not the case 2018-01-27 18:18:40 I was talking generally, I'm not using any of these, so cannot give any example, but I hope it exists 2018-01-27 18:20:10 honestly, if software like Atom uses AppImage or flatpak to distribute an official binary, I think we can just go without packaging nodejs-dependent software. 2018-01-27 18:20:28 package npm and nodejs itself? sure. applications, no. 2018-01-27 18:27:25 so you could say that people should simply download some flatpaks instead of looking for packages in distro. and it wouldn't be outrightly wrong to some extent, because flatpaks can be also seen (and sometimes probably are) as upstream saying: "I don't want to involve myself with all disto mess out there, just use what I use". the problem is we don't know what's in flatpaks, so distro cannot (or 2018-01-27 18:27:31 at least shouldn't, IMHO) recommend something that wasn't build by it. 2018-01-27 18:28:19 s/build by it/build by distro's builders/ 2018-01-27 18:29:29 well, keep in mind I'm only recommending this for desktop applications which use nodejs. 2018-01-27 19:31:31 hash update needed for main/busybox-initscripts 2018-01-27 19:44:41 done 2018-01-27 19:48:30 kaniini: thanks 2018-01-27 21:07:12 aerdan: i don't even think flatpak is an appropriate policy for adelie, much less alpine 2018-01-27 21:07:54 so much of the flatpaks out there have garbage quality assurance 2018-01-27 21:08:22 upstream developers generally don't know best in regards to integrating their software into a distribution 2018-01-27 21:08:56 it's a distribution model that works well for something like MacOS, but it's not workable for us 2018-01-27 21:09:38 AppImage and flatpak ship their own distribution inside the image 2018-01-27 21:09:57 which means if you expect things like locales to work cross-container-boundary, you may have a rude awakening :) 2018-01-27 21:17:22 is there a reason the SWAPSIZE env-var is no longer heeded by setup-alpine? this no longer does what I would expect: ROOTFS=btrfs BOOTFS=btrfs SWAPSIZE=0 setup-alpine 2018-01-27 21:28:20 why is it /usr/lib/lua5.3/liblua.a and not /usr/lib/liblua5.3.a ? 2018-01-27 21:30:22 hmm, i guess it's easier to shove a -L into a build system rather than renaming all -llua's 2018-01-28 08:07:02 The wide unicode -dev for ncurses wasn't packaged in. 2018-01-28 08:07:20 pamix uses it for emojis 2018-01-28 08:08:18 or linked to them indirectly via some UNICODE def 2018-01-28 10:15:27 any alpine dev online that can take a look and possibly merge this PR to eliminate a security issue? https://github.com/alpinelinux/aports/pull/3116 2018-01-28 10:16:06 clandmeter, Shiz around? ^ 2018-01-28 10:16:39 MITM due to improper input validation 2018-01-28 10:20:03 or ncopa for that sake 2018-01-28 14:25:09 ncopa: minified version of alpine-logo.svg by the way: https://hastebin.com/raw/atodohapoz 2018-01-28 14:25:11 functionally equal 2018-01-28 14:49:56 ncopa: did your designer friend reply on what font it is? 2018-01-28 14:56:23 looks like a modified Conthrax Extra Light 2018-01-28 19:56:35 Where would I submit a possible bug in an APKBUILD? I'm not sure if it is a bug or not. 2018-01-28 19:56:58 <_ikke_> bugs.a.o? 2018-01-28 19:57:03 just curious, what's the bug? also, ^ 2018-01-28 19:58:24 There's an explicit 'strip' command in jpegoptim that should be handled automatically by abuild 2018-01-28 19:58:54 all right, make an issue @ bugs.a.o and put assignee as the package maintainer 2018-01-28 19:58:58 "package update" category perhaps 2018-01-28 19:59:26 <_ikke_> make strip? 2018-01-28 19:59:34 <_ikke_> https://git.alpinelinux.org/cgit/aports/tree/community/jpegoptim/APKBUILD#n33 2018-01-28 19:59:52 Yes 2018-01-28 19:59:54 > options="!check" 2018-01-28 19:59:56 not !strp 2018-01-28 19:59:57 strip* 2018-01-28 20:00:46 Also, what is the current wisdom on explicitly setting builddir when it is the default ("$srcdir/$pkgname-$pkgver")? Is it a good idea to set builddir anyway or should it be unset? 2018-01-28 20:01:00 <_ikke_> unset or removed? 2018-01-28 20:01:16 I mean not declared in the APKBUILD 2018-01-28 20:01:19 <_ikke_> right 2018-01-28 20:01:25 I'm sorry 2018-01-28 20:01:27 <_ikke_> np 2018-01-28 20:01:47 <_ikke_> I'm not authority on this, but I'd say just leave it out 2018-01-28 20:03:08 What is your thought jirutka? You seem to be the main one reviewing patches and PRs so would you would know authoritively? 2018-01-28 20:22:55 There was a package in unmaintained/ (ipmiutil) that I fixed up and would be willing to maintain. But it was removed. Can I introudce it directly to testing/ or is there a procedure to follow? 2018-01-28 20:23:48 I would think introducing it directly to testing would be fine 2018-01-28 20:23:53 testing is, well, testing afterall 2018-01-28 20:32:24 I am so sorry for breaking things :( I had 'linux-headers' added locally to build-base because I was doing some experiments. 2018-01-28 20:45:22 I'm still getting this bug https://pastebin.com/hbWLGaa8 2018-01-28 20:46:56 Where is the APKBUILD? 2018-01-28 20:46:58 I removed it from the subpackages and it is still touching ycmd-doc*: Scanning shared objects 2018-01-28 20:47:30 I disabled the -doc subpackage. 2018-01-28 20:47:36 ERROR: Metadata for package ycmd-lic-20180123-r0 is too long. 2018-01-28 20:48:10 I need this fixed so I can upload the pavucontrol package 2018-01-28 20:48:42 I deleted that function 2018-01-28 20:48:44 The problem is the "-lic" subpackage. Note that when `abuild cleanpkg` is run, it only cleans the packages that would be built _now_. Try manually removing ycmd*-20180123-r0.apk from your /home/orson/packages/x86_64/ folder. 2018-01-28 20:48:47 it is cached 2018-01-28 20:49:00 s/folder/directory/ sorry 2018-01-28 20:49:22 (I used to do tech support for Macs so I learned to call them folders...) 2018-01-28 20:51:06 I use both 2018-01-28 21:03:07 What is the proper procedure when a package is broken on an architecture you don't have (in this case, testing/ladspa on aarch64). I might poke the package maintainer, but there is none :( Should I disable it on aarch64 in arch=? 2018-01-28 21:04:09 does apk's libfetch handle gzip or not? 2018-01-28 21:07:36 No. It doesn't have any support for Content-Encoding 2018-01-28 21:08:34 Any decent Web server will not send gzip-compressed HTTP responses without receiving Accept-Encoding though. 2018-01-28 21:08:49 Since the APK files themselves are already compressed, it doesn't make a lot of sense to support it. 2018-01-28 21:08:58 indeed - just checking 2018-01-28 21:28:11 I'm not sure if the flavour of libfetch was changed for the official apk release, but I think it did, and that flavour might have more features than the distro libfetch 2018-01-28 21:28:45 fabled knows this one best, I think it was he who made the change 2018-01-28 21:30:04 it's libfetch originally from netbsd, but with added features 2018-01-28 21:31:06 The question was "does apk's libfetch" so I looked at the actual libfetch/ dir in apk-tools.git :) 2018-01-28 21:31:23 AWilcox[m]: yeah, I am just a bit preoccupied 2018-01-28 21:31:30 didn't mean to ask a trivial question just to bother people 2018-01-28 21:38:58 Well, I already knew what I was looking for, since I am an ex-NetBSD dev 2018-01-28 21:39:08 No problem for me. 2018-01-28 21:42:12 I'm more familiar with openbsd and freebsd's code bases 2018-01-28 21:48:53 Should I back out the ipmiutil package until I have an answer? I don't want the aarch64 builder to be broken. 2018-01-28 21:49:01 Er, sorry, not ipmiutil. I meant ladspa 2018-01-28 22:28:05 hi 2018-01-28 22:28:22 hi you 2018-01-28 22:34:08 Is Alpine doing the split -openrc package thing? 2018-01-28 22:38:49 I'd love it to 2018-01-28 22:42:23 A. Wilcox: write to alpine-devel and try to get consensus. if not, we can decide if that will be a release goal for 3.8. 2018-01-28 22:51:58 I'd also like to mention that Qt5 Multimedia is currently using gstreamer 0.10 but it is fully compatible with gstreamer 1.0 and that is what we are using in Adelie. I don't want to change it without confirming this is ok 2018-01-28 23:18:01 seems fine to me 2018-01-29 00:04:37 i just pushed my redesign of `apk search` as the `apk list` applet. it can do a lot of useful things as a side effect of it's design 2018-01-29 00:05:07 it also has more obvious default behaviour verses apk search :) 2018-01-29 00:06:06 it also provides more details. so, for example, if you want to find out every license of every package on the system, you can do that with apk list and use an awk script to reformat it in a useful way :) 2018-01-29 00:11:41 are there plans to upgrade kernel versions on 3.7-stable? 2018-01-29 00:18:27 all currently supported distributions are being upgraded to 4.14 2018-01-29 00:18:38 and hardened is being dropped 2018-01-29 00:18:54 but not 100% of the work to drop hardened is done in edge yet 2018-01-29 00:18:59 so backports haven't been done 2018-01-29 00:19:10 virthardened too? 2018-01-29 00:19:29 yes, virthardened becomes linux-virt in edge 2018-01-29 00:19:59 that is nice 2018-01-29 00:20:03 we need help with getting hte last couple of hardened-only module APKBUILDs moved over 2018-01-29 00:20:16 then i will do the backports 2018-01-29 00:21:03 we do not want to use KAISER because in testing on edge branch, it was very unstable 2018-01-29 00:21:13 and the KPTI author said KAISER was flawed and exposed the kernel stack 2018-01-29 00:21:18 which seems even worse than meltdown 2018-01-29 00:21:55 I though 4.9.75 has KPTI? 2018-01-29 00:22:04 nope 2018-01-29 00:22:07 it has KAISER 2018-01-29 00:22:15 damn 2018-01-29 00:23:08 Greg K-H "helpfully" renamed KAISER to KPTI when he applied the KAISER patches, leading to this confusion 2018-01-29 00:23:25 but KPTI in 4.14 and "KPTI" in 4.9 are entirely different animals 2018-01-29 00:24:00 so that backports you were talking about are the real KPTI then? 2018-01-29 00:24:44 again, 4.4 and 4.9 do not have real KPTI, that have Greg K-H's bullshit KPTI that is actually KAISER. 2018-01-29 00:24:49 they* 2018-01-29 00:25:14 xentec: yes, we plan to backport real 4.14.15-r1 package from edge to all supported releases 2018-01-29 00:25:44 but the people with the release hats are waiting for the transition to complete in edge 2018-01-29 00:25:46 ah, so not only KPTI but the whole kernel version 2018-01-29 00:25:53 correct 2018-01-29 00:26:16 so if you want to help, make the remaining few kernel modules work on vanilla 4.14.15-r1 2018-01-29 00:26:17 ;) 2018-01-29 00:26:21 thanks for explaining, kaniini! 2018-01-29 00:26:39 if they are not updated soon we will just drop those modules 2018-01-29 00:26:45 and do the backport 2018-01-29 00:27:06 it should also be noted that KPTI does not protect against processes attacking other processes 2018-01-29 00:27:16 which, imo, is the more major problem 2018-01-29 00:28:14 and also the one that nobody is talking about, because it's terrifying and there's no fix other than a full CLFLUSH on every context switch 2018-01-29 00:28:26 and, well, if you think KPTI is bad 2018-01-29 00:28:30 that is a lot worse :) 2018-01-29 00:31:17 or exploitation across vmenter/vmexit 2018-01-29 00:33:08 x86 was a mistake. 2018-01-29 00:34:47 xentec: overall the situation of a broken patch that was rushed and completely untested being applied to an LTS kernel is new to us 2018-01-29 00:35:40 xentec: especially in a situation where that patch is entirely different and unrelated to what went mainline 2018-01-29 00:36:04 xentec: and itself opens security holes 2018-01-29 00:37:01 honestly Intel should pay everyone having to mitigate their fuck up 2018-01-29 00:37:56 that is indeed strange. until I though KPTI ist just a better name for KAISER but the code was the same 2018-01-29 00:38:03 until now* 2018-01-29 00:43:45 Aerdan: it's not just x86 2018-01-29 00:44:07 well 2018-01-29 00:44:18 meltdown is specific to nehalem and newer cpus 2018-01-29 00:44:28 they say 2018-01-29 00:44:58 but Spectre effects almost all modern cpus 2018-01-29 00:45:04 even POWER 2018-01-29 00:45:39 POWER9 does have steppings in production that do fix Spectre in hardware though 2018-01-29 00:45:46 but 8 is vulnerable 2018-01-29 00:47:12 The mitigation for Spectre is actually a lot simpler: realise your computer is a huge security hole and don't put anything in your computer that you don't want to shout in a restaurant. 2018-01-29 00:47:17 (So definitely don't use it for banking or such.) 2018-01-29 00:48:37 raptor computing systems erroneously announced power was not vulnerable to either which was funny 2018-01-29 00:48:50 it just turns out IBM was not in the embargo 2018-01-29 00:51:19 the big lesson here is that embargos 2018-01-29 00:51:21 are baf 2018-01-29 00:51:23 bad 2018-01-29 00:51:30 even 2018-01-29 00:52:54 and POWER4 as seen in the G5 powermac CPUs is vulnerable too 2018-01-29 00:53:56 http://tenfourfox.blogspot.com.au/2018/01/actual-field-testing-of-spectre-on.html 2018-01-29 00:59:16 one thing kaniini: KAISER exposing the kernel stack.. is there more you can give me a link to? 2018-01-29 01:06:31 xentec: from the kpti author himself: https://news.ycombinator.com/item?id=16087736 2018-01-29 01:08:46 thanks 2018-01-29 03:00:04 sorry, my mail to devel ML about licenses matters won't be on Sunday after all as I wrote earlier here, as drafting it takes a bit more time than I envisioned, and I'm already 4 hours into Monday in my timezone (CET), so have to continue tomorrow, and now is time for sleep. hopefully I'll finish it and send tomorrow before (or around) midnight in CET. 2018-01-29 09:00:43 hey there 2018-01-29 09:01:06 I'm still looking for someone who's able to merge this PR into the main aports repo: https://github.com/alpinelinux/aports/pull/3116 It's not a very high CVE that's assigned to this security issue but seeing that nobody seems to care for aports PRs with the T-security label assigned is not very good :( 2018-01-29 09:02:42 i’d like to contribute a package definition for a new package, cap’n proto (the quasi-successor to googles protobuf). am i right in that the procedure would be to first create an issue for the package request, and then … is a github PR accepted, or is a git-patch mail to the ML better? 2018-01-29 09:03:32 NebuK: I've had luck with a normal PR to github for new packages, see: https://github.com/alpinelinux/aports/pull/3045 2018-01-29 09:08:11 mh, lets try, then ^_^ 2018-01-29 09:12:42 i took care of 3116 2018-01-29 09:20:19 kaniini: what means "instead we import the patches manually"? 2018-01-29 09:25:20 i mean i understand that PRs on github aren't merged on github but instead commits are just manually copied off from github to the alpinelinux git system and the PR is then automatically closed, i do understand all of that, yea.. but this PR was seemingly ignored for a few days and nobody has actually copied the commit 2018-01-29 13:23:41 hi 2018-01-29 13:28:07 what are the semantics of the Contributor and Maintainer fields in an APKBUILD file? i couldn't find anything about that in the wiki 2018-01-29 13:28:44 firstname lastname , roughly. it's some RFC 2018-01-29 13:29:03 that's if I don't completely misremember the whole thing 2018-01-29 13:31:44 TBB: i meant relative to one another, i.e. what's the difference? 2018-01-29 13:33:53 Fusl: we used git even before github existed. GitHub mirror was added afterwards. We don’t depend on github, even if we have made it possible to submit patches via github. 2018-01-29 13:33:56 Importing manually means either curl url-to-patch or git pull url-to-pr 2018-01-29 13:33:58 We can ofcourse do rebase/fixup/squash, but our problem is that we are struggling keeping up 2018-01-29 13:34:01 And even security patches are drowning 2018-01-29 13:34:26 So it’s helpful if patches are easy for us to merge 2018-01-29 14:43:15 ncopa: how should I best handle XPTI fixes for Alpine 3.4? I have no problem preparing the change and build test it, but I have no HW running Alpine 3.4 bare-metal to test the result on. 2018-01-29 17:42:50 https://git.busybox.net/busybox/commit/?id=d8fd88a0915364c30769ec5c5a6b542517fd55f3 finally made it in 2018-01-29 17:43:58 kaniini: neat! that's not even a big diff :P 2018-01-29 17:45:16 alpine has been shipping with that patch for a while :P 2018-01-29 17:45:28 yeah, i remember you talking about it 2018-01-29 17:45:35 nice that it made its way into upstream 2018-01-29 17:53:44 oh 2018-01-29 17:53:57 it was always going to make its way upstream, i was not worried about it 2018-01-29 17:54:07 vda said he did not mind having it 2018-01-29 17:54:12 aha ok 2018-01-29 17:54:14 which is why i sent it upstream :) 2018-01-29 17:54:14 just took a long time? 2018-01-29 21:14:44 Okay. So we (Adelie) have 379 packages in main/ that we've touched. Obviously we won't be upstreaming all of them, but I do know that Alpine is happier with one package per PR. So I have a question: 2018-01-29 21:14:50 How many should we open at a time? Should I do, say, the first 25 or so, and then wait for those to be merged? 2018-01-29 21:14:58 I don't want to really spend this week opening > 300 PRs... 2018-01-29 21:15:09 And I'm sure that would be a major pain for Alpine as well. 2018-01-29 21:20:06 379... that sounds like fun 2018-01-29 21:25:14 I know... haha 2018-01-29 21:27:43 https://bpaste.net/raw/0b018c94a29f 2018-01-29 21:28:07 That doesn't include the patches, just APKBUILDs 2018-01-29 21:30:58 If I had to venture a guess: about 20 of those are going to be extra cross-compilation stuff that Alpine doesn't need, and another 30 or 40 will be PPC / big-endian fixes 2018-01-29 21:31:23 I don't know if Alpine is interested in merging in regular PowerPC (not 64le) support. 2018-01-29 21:52:07 from a contributor (to both projects) standpoint, it'd be less problematic to merge arch-specific fixes even if they're not relevant to 'official' archs, since that means no duplication of work is necessary should those archs become 'official' later. 2018-01-29 21:52:49 besides, for all I know the only reason Alpine doesn't officially do ppc64be is just that nobody's put the effort into making it happen yet. ;) 2018-01-29 21:58:31 if we get some build infrastructure and effort put into alpine on ppc64be, I can't see why it isn't possible 2018-01-29 22:11:58 How do I detect the current python implementations/versions for 2 and 3? in the APKBUILD? 2018-01-29 22:13:49 Does the package manager provide that info? 2018-01-29 22:14:54 This package doesn't have a setup.py file. 2018-01-29 22:16:52 You don't. Explicitly call python2 for 2.x and python3 for 3.x. 2018-01-29 22:17:09 Upstream officially deprecated /usr/bin/python as a thing and says nobody should use it. 2018-01-29 22:17:15 If a package uses it, that is a bug. Patch it and send it upstream. 2018-01-29 22:20:56 I am trying to install it to /usr/lib/python3.6/site-packages or whatever implementation/version of python 2018-01-29 22:21:21 on Gentoo they have a EPYTHON variable 2018-01-29 22:21:52 that spits out python3.6 or whatever 2018-01-29 22:22:36 cd "$builddir"; python(2|3) setup.py install 2018-01-29 22:22:46 Maybe tie it to Python version like kernel modules are tied to kernel version? 2018-01-29 22:22:54 s/; /\n/ 2018-01-29 22:22:55 danieli: they said the package doesn't have a setup.py file 2018-01-29 22:22:58 aha 2018-01-29 22:23:02 that's lame, not a proper setuptools package 2018-01-29 22:23:05 perhaps upstream should fix that 2018-01-29 22:52:34 I used the one provided by the blender package. 2018-01-29 22:53:42 I might go by letter 2018-01-29 22:53:53 Do all the "A" packages today, the "B" packages tomorrow, etc... 2018-01-29 22:53:59 I don't know how else to do this lol 2018-01-30 04:01:22 what could be causing this? Error relocating /usr/bin/pavucontrol: _ZTIN3Gtk8NotebookE: symbol not found 2018-01-30 04:03:05 are we using c++14 as the system default? 2018-01-30 04:07:48 rebuild pavucontrol? 2018-01-30 04:08:11 incidentally, if you're packaging LXQt, Adélie already has packages for that (for which you can blame me). 2018-01-30 06:40:05 I am working on making an electron package. 2018-01-30 06:40:36 to add on to the nodejs stack 2018-01-30 08:57:18 algitbot: ping 2018-01-30 08:57:36 <_ikke_> still asleep I guess :P 2018-01-30 08:57:43 :) 2018-01-30 08:58:00 \o/ 2018-01-30 09:00:16 algitbot: ping 2018-01-30 16:44:45 Hello, trying to run apache on alpine with mainline kernel, but it fails with the following error: 2018-01-30 16:44:48 [Thu Jan 01 01:01:28.116923 1970] [core:emerg] [pid 1108] (38)Function not implemented: AH00023: Couldn't create the mpm-accept mutex 2018-01-30 16:44:50 (38)Function not implemented: could not create accept mutex 2018-01-30 16:45:23 Do you know if it could come from linux build config? 2018-01-30 16:45:41 I tried to merge alpine vanilla’s config with mainline config, but the kernel just doesn’t boot 2018-01-30 16:47:25 <_ikke_> No actual idea, but perhaps? https://lists.debian.org/debian-apache/2003/11/msg00096.html 2018-01-30 16:54:34 Youbi: yes that seems like it could come from the build config 2018-01-30 16:54:56 i think you need to enable SysV IPC? 2018-01-30 16:57:37 Oh, checking 2018-01-30 16:58:42 _ikke_: thanks, but now that I think about it, on my desktop alpine with vanilla kernel, apache starts nicely 2018-01-30 16:58:53 So definitely kernel config related ^^ 2018-01-30 17:16:34 Shiz: checked my last config and it seems that SysV IPC was enabled… 2018-01-30 17:16:48 Building again, who knows ^^ 2018-01-30 17:17:38 CONFIG_SYSVIPC=y 2018-01-30 17:26:07 Well, built it again, and now it works :3 2018-01-30 17:48:50 ncopa: you might find this interesting:https://dl.alpinelinux.org/last-updated?mirrorstats 2018-01-30 17:52:29 heh 2018-01-30 18:28:09 does abuild have a linter? it didn't catch my syntax error: unterminated quoted string and complained about missing source= file 2018-01-30 18:28:57 the latter having higher precidence 2018-01-30 18:30:41 <_ikke_> koldbrutality: the shell is the linter 2018-01-30 18:34:11 I've suggested an apkbuild linter before 2018-01-30 18:34:22 Someone else said it was a good idea but I guess nobody actually wrote one 2018-01-30 18:34:50 Also, gst-plugins-base depends on liboil but orc replaces liboil and it builds fine without that dependency 2018-01-30 18:35:03 if I had a dollar for all the ideas everyone says are good and nobody moves their ass to implement... 2018-01-30 18:35:14 <_ikke_> Just use a regular shell linter 2018-01-30 18:35:29 <_ikke_> shellcheck for example 2018-01-30 18:53:19 there's no package for shellcheck 2018-01-30 18:54:20 <_ikke_> hmm, I'll check if I can package it then 2018-01-30 18:54:25 shouldn't be hard 2018-01-30 18:54:30 <_ikke_> it's haskell 2018-01-30 18:54:37 /shrug 2018-01-30 18:55:01 grab some inspiration from here https://git.archlinux.org/svntogit/community.git/tree/trunk/PKGBUILD?h=packages/shellcheck 2018-01-30 18:55:16 <_ikke_> yeah, that's what I usually do 2018-01-30 18:55:22 :^) 2018-01-30 18:55:40 if there's a nice package for it in arch, why not steal some ideas from there? 2018-01-30 18:55:47 and void, if they have some patches we need 2018-01-30 21:34:46 can we get this upgrade in? https://github.com/alpinelinux/aports/pull/3085 2018-01-30 21:37:25 kaniini: ^ 2018-01-30 21:38:06 i'm deep in x86 context switch code right now ( https://github.com/kaniini/libucontext ) so we can have the makecontext & friends functions in alpine 2018-01-30 21:38:11 all right 2018-01-30 21:38:19 _ikke_: you have push access, don't you? 2018-01-30 21:39:01 <_ikke_> To where? 2018-01-30 21:39:08 aports 2018-01-30 21:39:12 see that github PR 2018-01-30 21:39:54 he does not 2018-01-30 21:40:03 all right 2018-01-30 21:40:09 tag someone who does, if you don't mins 2018-01-30 21:40:14 s/mins/mind/ 2018-01-30 21:41:31 Shiz: you, perhaps? 2018-01-30 21:41:50 i can't push to main/ 2018-01-30 21:42:50 hmm all right 2018-01-30 21:43:02 beyond that, i believe that 2018-01-30 21:43:05 - security team should handle T-Security PRs 2018-01-30 21:43:13 there is a security team? 2018-01-30 21:43:17 - i should never touch anything related to php 2018-01-30 21:43:19 I thought it wasn't established yet 2018-01-30 21:43:22 lol 2018-01-30 21:43:25 there is not 2018-01-30 21:43:31 i am just saying, that is what i believe 2018-01-30 21:43:37 when one is established, agreed 2018-01-30 21:43:49 I'll review it. I'd push it if I could 2018-01-30 21:45:42 re that PR: it's a trivial version bump, plus removing a patch I assume no longer is needed. it should be safe to push. 2018-01-30 22:30:05 _ikke_: regarding shellcheck, here are the dependencies: https://bugs.alpinelinux.org/issues/7105 2018-01-30 22:30:12 also +1 for linting 2018-01-30 22:31:43 shfmt may make sense to automatically format APKBUILDs if that is desired (it supports posix shell). it's written in go, which probably makes it challenging to package as well. https://github.com/mvdan/sh 2018-01-30 22:36:00 in other news, I finally made that abuild patch, which makes "-f" only rebuild up-to-date packages, and not skip checksums: https://github.com/alpinelinux/abuild/pull/31 2018-01-30 22:36:23 go is actually okay for packaging. it's *rust* that's obnoxious. 2018-01-30 22:36:46 uhhuh? 2018-01-30 22:36:47 they're both awful ;) 2018-01-30 22:37:02 ^ 2018-01-30 22:37:34 well, the argument that you need rust to build rust is off the table at least 2018-01-30 22:39:09 mrustc is making good progress toward fixing that 2018-01-30 22:40:18 it's already "Capable of building a fully-working copy of rustc, but not yet suitable for everyday use." 2018-01-30 22:40:18 from their README.md 2018-01-30 23:49:39 how much of Arch Linux's build feature is incorporated in Alpine? 2018-01-30 23:49:52 can we use like the git as a source? 2018-01-30 23:50:38 as in PKGBUILD vs APKBUILD 2018-01-30 23:51:27 <_ikke_> koldbrutality: abuild is an independent implementation 2018-01-30 23:51:46 the formats are similar, but entirely different things 2018-01-30 23:53:11 chromium.googlesource.com doesn't have the as download .zip or i can't find it. 2018-01-30 23:53:30 <_ikke_> koldbrutality: what about git archive? 2018-01-30 23:53:58 download source up to git commit as a zip. 2018-01-30 23:54:27 <_ikke_> koldbrutality: I think it should be possible, trying to find a reference 2018-01-30 23:56:01 <_ikke_> https://wiki.alpinelinux.org/wiki/APKBUILD_examples:Git_checkout 2018-01-30 23:57:08 there is also a similar function in abuild. abuild snapshot 2018-01-31 00:16:38 clandmeter: any idea if there's a reason patchwork depends on both python2 and python3? 2018-01-31 00:17:33 i have no idea 2018-01-31 00:18:49 I'll check it out 2018-01-31 00:20:28 the apkbuild had this? 2018-01-31 00:25:48 clandmeter: yes, I merely modified the old one 2018-01-31 00:30:14 it's most likely because of py-django-registration that depends on python2 2018-01-31 00:31:21 aha, thank you 2018-01-31 00:31:45 I'll have another look at it tomorrow to familiarize myself a bit more in-depth with it 2018-01-31 00:32:56 I'm not a python expert (not even a newbie), so maybe this registration can be very easily fixed to be build for py3 2018-01-31 00:33:41 yeah, probably. I'm good with py, so I'll check tomorrow 2018-01-31 00:33:44 good night :) 2018-01-31 00:33:57 yeah, I should go to sleep too ;) 2018-01-31 00:35:13 (or maybe it's APKBUILD thing only and no patches will be even needed. I'm not familiar with building python stuff in AL.) 2018-01-31 00:38:47 clandmeter: out of curiosity, why patchwork has been downgraded to community? 2018-01-31 00:39:35 is it a downgrade? 2018-01-31 00:39:59 i see no reason it should be in main, thats why i moved it. 2018-01-31 00:40:22 main should have a limited set of packages 2018-01-31 00:41:10 I'm not denying it, but it would be good to state reason behind promotions or demotion in commit message 2018-01-31 00:42:23 there is no specific reason, its just that some packages are still in main since we introduced community. 2018-01-31 00:42:44 i could have added it, but in this case it was kind of obvious (at least to me). 2018-01-31 00:43:10 moving things to main always gets a reason in my commit msg. 2018-01-31 00:44:09 there isn't any rule that what's used by alpine infra should be in main? 2018-01-31 00:44:34 main pkgs should support the full 2 years of a release 2018-01-31 00:44:53 I know that, that's not what my question was 2018-01-31 00:44:54 no i dont think that is a reason to have it in main. 2018-01-31 00:44:58 ok 2018-01-31 00:45:36 and i dont think we ever mentioned such rule. 2018-01-31 00:46:10 I'm not implying there was such, just asked (because such rule wouldn't be unreasonable). 2018-01-31 00:46:24 in a perfect world we should upgrade our infra to a new version when it is availalbe 2018-01-31 00:46:59 I won't deny that either. 2018-01-31 00:47:40 if you can provide me a good reason to have infra related packages in main, im sure i will think about it. 2018-01-31 00:51:34 ok its time for me to get some sleep. you can leave me a msg if you have some insights, or drop me a mail. 2018-01-31 00:51:41 if AL had good documentation covering also its infra setup, then having infra-related packages in main would be useful for AL spin-offs not needing (or even not being able to) to upgrade to newer AL every half year. 2018-01-31 00:52:02 but as it's not the case yet, then it's not a good enough argument 2018-01-31 00:52:07 for now 2018-01-31 00:57:27 ACTION looks up. 2018-01-31 00:57:54 AIUI, main/ is intended for things directly supported by Alpine; community/ is intended for things supported by the community at large. 2018-01-31 00:58:27 things being moved to community/ from main/ is an indication that they won't be supported by Alpine Linux core team. 2018-01-31 01:00:56 there is also 2 years vs 6 months support thing, and if upstream doesn't provide such longer support, package usually won't be promoted to main/ unless it's seemed as important enough to burden AL devs with supporting on their own. 2018-01-31 01:01:38 Aerdan: i think that definition is incorrect, but until its been defined and written down we can discus this forever. 2018-01-31 01:02:10 the only definition for now is how przemoc has just stated. 2018-01-31 01:02:45 there has also been talk to have pkgs in main only to bring up the system. 2018-01-31 01:03:22 bring up as build, boot and get into the shell, I presume? 2018-01-31 01:03:41 i think that was the logic behind it yes. 2018-01-31 01:04:18 we also had a lot of discussion about the testing repo naming 2018-01-31 01:04:33 people took it too literally, which can be confusing. 2018-01-31 01:04:41 just as you could with community. 2018-01-31 01:05:34 core/main/community/staging would also be possible 2018-01-31 01:05:42 main/ could be split into core/ and main/, where core/ would be having 2 years (4 releases), but main/ for instance only 1 year (2 releases) and community would remain half a year (1 release) 2018-01-31 01:05:50 just what I thought 2018-01-31 01:07:00 ok now i really need to hit my bed, 2am+ here 2018-01-31 01:07:06 same here 2018-01-31 01:07:09 gnite! 2018-01-31 01:07:15 nite! 2018-01-31 02:19:42 I prefer rolling release where everything is (near) bleeding edge. If people want to have LTR, they should just fork alpine as alpine-ltr or switch to something like ubuntu. If the release cycle is slow, then you won't get features and the libraries are old and has exploits. 2018-01-31 02:20:33 ltr (long term relationship) -> lts (long term support) 2018-01-31 02:25:44 If the linux kernel was in core and release cycle was long then people would have meltdown for long ass time. 2018-01-31 02:33:26 since alpine doesn't have a slot system. you could probably do something like the qt4 qt5 thing but add -lts packages 2018-01-31 02:34:15 if you want rolling release, stick to edge. 2018-01-31 03:02:53 koldbrutality: that is what edge is for (rolling release) 2018-01-31 05:48:04 For systems that must Just Work(tm), rolling releases are a pain and lead to 'rather not update, in case it breaks something else again'. 2018-01-31 06:58:59 detha: that's why alpine makes you have to choose to use rolling release :D 2018-01-31 07:01:58 Xe: which is a good thing. I was just commenting on the 'I prefer rolling release...' earlier. 2018-01-31 10:49:45 how long is alpine going to stick with gcc 6.x? 2018-01-31 10:53:23 until there's good reason to use something newer. 2018-01-31 10:53:39 gcc has this problem where newer isn't always strictly better. 2018-01-31 11:02:57 c++17 support and retpoline may be one of those reasons 2018-01-31 11:04:12 retpoline can be backported. 2018-01-31 13:07:32 I need to update qt5 to 5.10 to fix gns3 2018-01-31 13:10:07 ollieparanoid[m]: ^ 2018-01-31 13:12:10 adelie will be working on that just as soon as we can reduce the deltas between adelie's aports and alpine's. 2018-01-31 13:12:17 (I think, anyway.) 2018-01-31 13:47:14 Aerdan[m], does it mean that I should wait? 2018-01-31 13:47:38 I don't want to create troubles to you, so I believe we can coordinate the efforts for that 2018-01-31 13:47:55 as ncopa said, postmarketos wants notices when qt is upgraded 2018-01-31 13:47:59 i believe it was qt 2018-01-31 13:48:09 danieli, correct 2018-01-31 13:48:33 that's why I'm asking here. 2018-01-31 13:48:37 aha 2018-01-31 13:48:57 Rather than pushing the upgrade I can send a patch to GH 2018-01-31 13:49:02 as PR 2018-01-31 13:49:20 that works I suppose, it'll leave some time for review and remarks 2018-01-31 13:49:25 so postmarketos is notified...hopefully :) 2018-01-31 13:49:41 if not, they gotta pay a bit of attention :P 2018-01-31 13:49:48 fcolista: best person to ask is AWilcox. 2018-01-31 13:50:04 but I believe we can find a way to cooperate with that also 2018-01-31 13:50:09 AWilcox[m]: ping ding 2018-01-31 13:50:10 indeed 2018-01-31 13:50:52 if there are "sensitive" packages for postmarketos, like qt, i think we should think to a procedure to communicate that 2018-01-31 13:51:39 'cause the problem is: $maintainer is notified via email that new $package version is available 2018-01-31 13:51:49 $maintainer push the upgrade 2018-01-31 13:52:24 if $package is sensitive for postmarketos/adelie/whatever_we_might_have_in_the_future 2018-01-31 13:52:29 then we can create a problem 2018-01-31 17:05:04 fcolista: you're free to push qt 5.10, pmos has no issues with it, and I don't think adelie does either 2018-01-31 17:07:57 in this case Qt 5.10 might resolve some issues for us. if not, well it won't hurt either 2018-01-31 17:16:00 PureTryOut[m]: have you tested 5.10? 2018-01-31 17:16:31 I haven't no, I just roll with upstream most of the time. upstream in this case is Alpine 2018-01-31 17:17:15 but right now we shouldn't be afraid to break much, as pmOS isn't yet ready to use as a daily driver for anyone anyway 2018-01-31 17:17:27 mm, that's a good point 2018-01-31 17:17:35 I have some questions, I'll ask in your own channel 2018-01-31 17:17:40 still glad you asked us though 2018-01-31 18:33:09 fcolista: we prefer to stay on Qt 5.9 LTS 2018-01-31 18:38:08 fcolista: what is the actual problem with GNS3 2018-01-31 18:38:39 latest mainline kernel is 4.15, latest stable 4.14.16 2018-01-31 18:39:49 to be very clear, adelie's position on Qt is that we want to always use the LTS. 2018-01-31 18:40:02 non-LTS Qt branches get basically no QA 2018-01-31 18:43:39 indeed 2018-01-31 18:46:59 well 2018-01-31 18:47:07 you said we had no issue with it ;) 2018-01-31 18:47:44 Qt 5.9 is definitely something that needs to stay put. there are some issues with multihead setups in 5.10 2018-01-31 18:48:03 also memory leaks 2018-01-31 18:48:21 (we tried Qt 5.10 already, it is definitely not something we want to upgrade to) 2018-01-31 18:49:10 I said I don't think you do 2018-01-31 18:49:12 I was obviously wrong 2018-01-31 18:50:05 take it with a pinch of salt 2018-01-31 18:51:01 Oh, 4.14.16 is out? That'll be a good test of the new kernel stuff I'm working on 2018-01-31 18:51:22 Trying to make it so upgrades don't pull modules out from under the running kernel 2018-01-31 18:51:44 it is indeed out :) 2018-01-31 18:51:58 out today 2018-01-31 18:52:29 same with 4.9.79 LTS and 4.4.114 LTS 2018-01-31 18:53:12 unless those 4.9 and 4.4 releases remove KAISER in favour of current KPTI, who cares 2018-01-31 18:53:21 ikr 2018-01-31 18:53:49 (not that KPTI is really a good solution, but...) 2018-01-31 19:12:07 i still say they should have called it User Address Space Separation (UASS) 2018-01-31 19:12:27 dat UASS 2018-01-31 19:13:49 nah, they shoulda stuck with FUCKWIT 2018-01-31 19:14:11 FUCKWIT isn't media friendly enough I guess ;) 2018-01-31 19:59:05 Once patch that I have held in my queue is our apk-tools patch to build against OpenSSL. It doesn't right (it uses a libressl API). Is that compatibility desired? 2018-01-31 20:00:46 yes 2018-01-31 20:00:48 send it to me i will put it in 2018-01-31 20:04:00 kaniini: http://foxkit.us/linux/0001-libfetch-support-OpenSSL.patch 2018-01-31 20:04:16 kaniini: please test it on LibreSSL first, I have not 2018-01-31 20:04:29 Though I do expect it will be fine 2018-01-31 20:05:49 it built and runs fine 2018-01-31 20:05:50 i pushed it 2018-01-31 20:06:58 can people test http://github.com/kaniini/libucontext on their 32-bit ARM devices 2018-01-31 20:07:02 make ARCH=arm check 2018-01-31 20:07:12 as long as the test program does not crash, it's good 2018-01-31 20:07:36 i encountered some processor errata on certain devices while writing the ARM implementation 2018-01-31 20:07:44 so i want to be certain there is not more :| 2018-01-31 20:11:07 kaniini: works great on my ARMv6 Raspberry Pi 1B+ (/proc/cpuinfo at https://bpaste.net/raw/eb6ab0c1462d - it's a BCM2708) in both Debian Jessie and Alpine 3.7 2018-01-31 20:11:28 No crashes, "start f2\nstart f1\nfinish f2\nfinish f1\n" is the output 2018-01-31 20:14:29 i tested on ARMv6 marvell sheevaplug IoT device 2018-01-31 20:14:34 and a scaleway server 2018-01-31 20:14:46 and ARMv7 raspberry pi 2 2018-01-31 20:15:06 the sheevaplug had the errata 2018-01-31 20:15:37 specifically, restoring the context by hand was causing it to loop indefinitely 2018-01-31 20:15:55 i suspect faulty branch prediction logic 2018-01-31 22:38:53 fcolista, danieli: thanks for the heads up about the upgrade! as PureTryOut noted we're far from daily driver level - when we get somewhat close to it, then yes, setting up a defined process to communicate such updates would make sense. The good part is, that we have a testcase set up, which checks if our qt5 packages, which are not upstreamed yet (i.e. qt5-qtwebengine) have the same version as Alpine's 2018-01-31 22:38:54 qt5-qtbase. So we would notice it pretty soon and could kick off our build then. So this test case is good enough for the current situation we have. -- regarding Adelie, I think they don't use Alpine's binary repositories and build all packages on their own. So they would not be affected by binary incompatibilities in Alpine's repos, right? 2018-01-31 22:51:23 The eventual goal is to be able to offer Alpine's community repo without changes 2018-01-31 22:51:53 That would therefore be a major change 2018-01-31 23:19:03 we desire to be binary-compatible with alpine repositories, yes